# 107 link-building statistics, explained for SEOs SEOlutions · Original sources checked 9 September 2026 This reference contains 107 distinct findings from 33 studies and research releases. Comparisons count once. Research periods vary; a 2026 update does not turn older measurements into 2026 data. The plain-language explanations and action steps are SEOlutions’ interpretation. Survey answers describe respondents, observational relationships do not establish causation, and individual experiments are not universal forecasts. Read the portal article: https://pr.seolutions.biz/blog/link-building-statistics-insights/ ## Links and rankings See how links relate to Google rankings, and why links are only one part of the picture. ### 1. Links and rankings have a modest relationship **0.255 / 0.248 correlation** — In Ahrefs’ January 2025 US keyword analysis, referring-domain count had a 0.255 rank correlation and backlink count 0.248. **In plain English:** Pages with more links tended to rank a little higher, but the connection was weak. Other factors clearly matter too. **What you can do:** 1. Compare your page with the pages that already rank for one important search. 2. Check the content and internal links before deciding that more backlinks are needed. [Ahrefs: Google Says “Links Matter Less”—We Looked at 1,000,000 SERPs to See if It’s True](https://ahrefs.com/blog/links-matter-less-but-still-matter/) - Evidence: Website study; 2025 release · observation dates unstated - Sample: 1,000,000 keywords - Population: Top 20 Google results for the one million highest-volume US keywords - Data period: not reported; published January 2025 - Source location: Correlation table; introduction and methodology - Limit: Observational Spearman correlations; high-volume US queries are not representative of all searches. Correlation cannot establish ranking causation. - Evidence ID: S001; checked 2026-09-09 ### 2. Most indexed pages had no estimated traffic **96.55% of pages** — Ahrefs’ 2023 analysis estimated zero Google organic traffic for 96.55% of roughly 14 billion indexed pages. **In plain English:** Most pages in the study received no estimated Google visits. Publishing a page alone does not bring it an audience. **What you can do:** 1. Use Search Console to find pages that receive no views or clicks. 2. Improve or remove pages with no clear audience or purpose before promoting them. [Ahrefs: 96.55% of Content Gets No Traffic From Google. Here’s How to Be in the Other 3.45%](https://ahrefs.com/blog/search-traffic-study/) - Evidence: Website study; 2023 release · traffic estimates - Sample: approximately 14 billion pages; separate subset of about 20 million pages with no referring domains - Population: Pages in the Ahrefs index, with estimated Google organic traffic - Data period: not reported; 2023 report - Source location: Opening findings; study limitations - Limit: Traffic is estimated from Ahrefs keyword coverage, not measured analytics. Indexed sample is quality-biased and does not represent every page on the web. - Evidence ID: S002; checked 2026-09-09 ### 3. Some unlinked pages still attracted search visits **2,997 of about 20 million** — In a separate subset in Ahrefs’ 2023 traffic study, 2,997 of about 20 million pages with no referring domains exceeded 1,000 estimated monthly Google visits. **In plain English:** A small number of pages got search visits even though Ahrefs found no links from other sites. Links are not always required. **What you can do:** 1. Find pages that get search visits despite having few links from other websites. 2. Learn from their content and internal links before seeking more backlinks. [Ahrefs: 96.55% of Content Gets No Traffic From Google. Here’s How to Be in the Other 3.45%](https://ahrefs.com/blog/search-traffic-study/) - Evidence: Website study; 2023 release · traffic estimates - Sample: approximately 14 billion pages; separate subset of about 20 million pages with no referring domains - Population: Pages in the Ahrefs index, with estimated Google organic traffic - Data period: not reported; 2023 report - Source location: Section discussing pages without backlinks - Limit: Traffic is estimated from Ahrefs keyword coverage, not measured analytics. Indexed sample is quality-biased and does not represent every page on the web. - Evidence ID: S003; checked 2026-09-09 ### 4. Domain authority is only one clue **0.131 correlation** — Domain Rating had a 0.131 ranking correlation in Ahrefs’ 2025 high-volume US keyword sample. **In plain English:** A strong website score did not closely match higher rankings. It is one clue, not a final verdict. **What you can do:** 1. Check topic fit, real traffic, article quality and outgoing links as well as DR. 2. Read five recent articles before approving the website. [Ahrefs: Google Says “Links Matter Less”—We Looked at 1,000,000 SERPs to See if It’s True](https://ahrefs.com/blog/links-matter-less-but-still-matter/) - Evidence: Website study; 2025 release · observation dates unstated - Sample: 1,000,000 keywords - Population: Top 20 Google results for the one million highest-volume US keywords - Data period: not reported; published January 2025 - Source location: Correlation table: DR - Limit: Observational Spearman correlations; high-volume US queries are not representative of all searches. Correlation cannot establish ranking causation. - Evidence ID: S083; checked 2026-09-09 ### 5. Local queries had a stronger link relationship **0.33 correlation** — For local queries in Ahrefs’ 2025 analysis, backlinks and referring domains each had a 0.33 ranking correlation. **In plain English:** Links and local rankings moved together more closely than in the full sample. Local search markets may behave differently. **What you can do:** 1. Compare linking websites for each important local search and location. 2. Repeat the check when local competitors or search results change. [Ahrefs: Google Says “Links Matter Less”—We Looked at 1,000,000 SERPs to See if It’s True](https://ahrefs.com/blog/links-matter-less-but-still-matter/) - Evidence: Website study; 2025 release · observation dates unstated - Sample: Local-query subset of one million keywords; subset size undisclosed - Population: Top-20 results for the local queries in the parent US keyword sample - Data period: not reported; published January 2025 - Source location: Local and non-local query comparison - Limit: Observational Spearman correlations; high-volume US queries are not representative of all searches. Correlation cannot establish ranking causation. The local-query subset size is not disclosed. - Evidence ID: S085; checked 2026-09-09 ## Choosing worthwhile placements Learn how to judge a website and the exact page where your link would appear. ### 6. Reciprocal linking was common in the sample **73.6% of domains** — In Ahrefs’ historical study of 140,592 domains with search traffic, 73.6% had at least one reciprocal linking relationship. **In plain English:** Many websites linked to sites that also linked back. This can happen naturally between related publishers. **What you can do:** 1. Check whether each two-way link is genuinely helpful to readers. 2. Review repeated link swaps between the same small group of websites. [Ahrefs: Reciprocal Links: Do They Help or Hurt Your SEO?](https://ahrefs.com/blog/reciprocal-links/) - Evidence: Website study; Historical study · dates unstated - Sample: 140,592 domains - Population: Domains with search traffic; followed incoming and outgoing links only - Data period: not reported - Source location: How common are reciprocal links on the web? - Limit: Historical study; page updated March 11, 2020, but original publication and fieldwork dates are not established. Traffic-selected sample has survivor bias. Reciprocity is not proof of an exchange scheme or its safety. - Evidence ID: S007; checked 2026-09-09 ### 7. Substantial two-way overlap also appeared **27.4% of domains** — In the same undated Ahrefs sample, 27.4% of domains had at least 15% overlap between the domains they linked to and those linking back. **In plain English:** More than a quarter of the sites had a lot of two-way links. The number describes what existed, not what you should copy. **What you can do:** 1. Find which websites exchange the most links with your site. 2. Keep natural partnerships, but do not aim for this study’s percentage. [Ahrefs: Reciprocal Links: Do They Help or Hurt Your SEO?](https://ahrefs.com/blog/reciprocal-links/) - Evidence: Website study; Historical study · dates unstated - Sample: 140,592 domains - Population: Domains with search traffic; followed incoming and outgoing links only - Data period: not reported - Source location: Distribution of reciprocal-link overlap - Limit: Historical study; page updated March 11, 2020, but original publication and fieldwork dates are not established. Traffic-selected sample has survivor bias. Reciprocity is not proof of an exchange scheme or its safety. - Evidence ID: S008; checked 2026-09-09 ### 8. Reciprocal links appear on ranking pages **43.7% of pages** — Ahrefs found reciprocal links on 43.7% of 112,440 top-ranking pages in a separate historical sample. **In plain English:** Two-way links also appeared on high-ranking pages. That does not prove that planned link swaps are helpful or safe. **What you can do:** 1. Record why each site links to the other and how it helps readers. 2. Review repeated swaps, copied link text or links to unrelated pages. [Ahrefs: Reciprocal Links: Do They Help or Hurt Your SEO?](https://ahrefs.com/blog/reciprocal-links/) - Evidence: Website study; Historical study · dates unstated - Sample: 112,440 top-ranking pages from about 10,000 selected queries - Population: Top-ranking pages for non-branded queries with keyword difficulty 40–60 - Data period: not reported - Source location: How much reciprocal linking happens in the top 10? - Limit: Historical study; page updated March 11, 2020, but original publication and fieldwork dates are not established. Traffic-selected sample has survivor bias. Reciprocity is not proof of an exchange scheme or its safety. This subset used about 10,000 non-branded queries with keyword difficulty 40–60; it does not establish that exchange schemes are safe. - Evidence ID: S086; checked 2026-09-09 ### 9. Many listings had little estimated traffic **39.6% zero; 78% below 1,000** — In BuzzStream’s 2026 marketplace sample, 39.6% of sites had zero estimated organic traffic; 78% had fewer than 1,000 monthly visits. **In plain English:** Many sites offered for paid placement had little or no estimated Google traffic. A large marketplace does not guarantee a real audience. **What you can do:** 1. Remove marketplace sites with little estimated traffic, then check the rest by hand. 2. Record the tool and date because traffic estimates can be wrong. [BuzzStream: Cost of Guest Posts, Based on 500k Sites [Updated 2026 Data]](https://www.buzzstream.com/blog/guest-post-costs/) - Evidence: Marketplace study; Jun 2026 release · marketplace listings - Sample: 257,267 sites with price, Domain Rating and estimated traffic data - Population: Complete listings in one unnamed link marketplace - Data period: June 2026 release; collection dates undisclosed - Source location: Key takeaways: traffic distribution - Limit: One marketplace, not the whole web. Listed quotes are not completed transactions. Traffic is an SEO-tool estimate; incomplete listings were excluded. Both traffic bands describe the same sample and count as one distribution finding. - Evidence ID: S104; checked 2026-09-09 ### 10. Few listings passed the study’s top tiers **1.37% of listed sites** — Only 1.37% of sites in BuzzStream’s 2026 sample met its “High Quality” or “Top Tier” metric thresholds. **In plain English:** Very few listings met this study’s strictest quality rules. Always ask what a seller’s quality label actually means. **What you can do:** 1. Ask sellers exactly what labels such as “premium” mean. 2. Judge every site using your own relevance, quality and audience checks. [BuzzStream: Cost of Guest Posts, Based on 500k Sites [Updated 2026 Data]](https://www.buzzstream.com/blog/guest-post-costs/) - Evidence: Marketplace study; Jun 2026 release · marketplace listings - Sample: 257,267 sites with price, Domain Rating and estimated traffic data - Population: Complete listings in one unnamed link marketplace - Data period: June 2026 release; collection dates undisclosed - Source location: Quality tier distribution; threshold definitions - Limit: One marketplace, not the whole web. Listed quotes are not completed transactions. Traffic is an SEO-tool estimate; incomplete listings were excluded. These are the author’s DR/traffic tiers, not an editorial audit: DR 71–80 with 50,000+ traffic, or DR 81+ with 100,000+ traffic. - Evidence ID: S105; checked 2026-09-09 ### 11. Niche labels do not establish an audience **70% of gambling listings** — About 70% of roughly 10,000 gambling sites in BuzzStream’s 2026 marketplace subset had zero estimated organic traffic. **In plain English:** Most gambling-category sites in this marketplace had no estimated search traffic. A niche label can be misleading. **What you can do:** 1. Read recent articles and check search visibility before trusting a niche label. 2. Remove sites whose real content or audience does not match your campaign. [BuzzStream: Cost of Guest Posts, Based on 500k Sites [Updated 2026 Data]](https://www.buzzstream.com/blog/guest-post-costs/) - Evidence: Marketplace study; Jun 2026 release · marketplace listings - Sample: Roughly 10,000 gambling-site listings - Population: Gambling-category subset of one marketplace’s complete listings - Data period: June 2026 release; collection dates undisclosed - Source location: Gambling deep dive - Limit: One marketplace, not the whole web. Listed quotes are not completed transactions. Traffic is an SEO-tool estimate; incomplete listings were excluded. A subset of one marketplace; not a statement about all gambling publishers. - Evidence ID: S106; checked 2026-09-09 ### 12. DR was a popular screening metric **67% DR; 42% DA** — In Aira’s 2022 survey, 67% used Domain Rating and 42% used Domain Authority to assess link quality. **In plain English:** DR was a popular first check, but it cannot show whether a specific article or audience is a good fit. **What you can do:** 1. Use DR only to create a first list, then inspect each page. 2. Write down why you accepted or rejected each website. [Aira: 11 Takeaways from the 2022 State of Link Building Report](https://aira.net/blog/11-takeaways-2022-state-of-link-building-report/) - Evidence: Survey; Mar–May 2022 · historical survey - Sample: 270 respondents; branching questions have different denominators - Population: Self-selected SEO professionals answering the annual link-building survey - Data period: 16 March–2 May 2022 - Source location: Takeaway 2; report metric question - Limit: Historical self-reported practice and opinions. Multiple-choice questions may allow several answers; responses are not measured ranking effects. - Evidence ID: S115; checked 2026-09-09 ### 13. Page relevance and domain relevance differ **62% vs 52%** — Aira’s 2022 respondents assessed relevance using the linking page’s topic (62%) and the domain’s topic (52%). **In plain English:** A website can cover your topic while the actual placement page does not. You need to check both. **What you can do:** 1. Check whether both the website and the planned article fit your topic. 2. Choose the page that helps readers most, even on a broader website. [Aira: 11 Takeaways from the 2022 State of Link Building Report](https://aira.net/blog/11-takeaways-2022-state-of-link-building-report/) - Evidence: Survey; Mar–May 2022 · historical survey - Sample: 270 respondents; branching questions have different denominators - Population: Self-selected SEO professionals answering the annual link-building survey - Data period: 16 March–2 May 2022 - Source location: Takeaway 7; relevance criteria - Limit: Historical self-reported practice and opinions. Multiple-choice questions may allow several answers; responses are not measured ranking effects. - Evidence ID: S117; checked 2026-09-09 ### 14. A strong domain score can hide a weak page score **Average DR 61; UR 11** — Reboot reported an average DR of 61 and UR of 11 for its 2024 digital PR coverage. **In plain English:** A high domain score can hide a weak page. The page matters because that is where people will see the link. **What you can do:** 1. Check the exact page’s traffic, internal links and surrounding content. 2. Reject a weak page even when the website has a strong score. [Reboot Online: Digital PR Statistics](https://www.rebootonline.com/digital-pr-statistics/) - Evidence: Website study; 2024 coverage · one agency - Sample: More than 6,200 links acquired; exact metric-specific denominator undisclosed - Population: Reboot’s own digital PR coverage - Data period: Agency coverage acquired during 2024 - Source location: Quality of backlinks: 2024 DR and UR - Limit: One agency’s selected campaign mix. DR and UR are third-party metrics, not Google scores. The source page also contains unrelated secondary statistics. - Evidence ID: S120; checked 2026-09-09 ## Costs and budgets Understand what a price includes before you compare providers. ### 15. Budget approval was not easy for most **29% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 29% said securing a link-building budget was very or somewhat easy. **In plain English:** Only a minority said budget approval was easy. Clear goals and costs make the request easier to judge. **What you can do:** 1. Explain which pages, audience and campaign idea the budget will support. 2. Start with a small test budget, then agree when to stop or spend more. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Link Building Budgeting; ease of securing budget - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S021; checked 2026-09-09 ### 16. Many teams did not know cost per link **19.6% at $300–$500; 39.2% unsure** — In BuzzStream’s 2026 digital PR survey, 19.6% reported average link costs of $300–$500, while 39.2% did not know their cost per link. **In plain English:** Many respondents did not know their cost per link. The number is unreliable unless everyone counts cost in the same way. **What you can do:** 1. Count agency fees, tools, production and staff time in the total cost. 2. Check the cost again after removed or rejected links are excluded. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Digital PR Budgeting, question 3 - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S024; checked 2026-09-09 ### 17. A retainer includes more than a link **US$5,458 per month** — BuzzStream’s 2025 survey of PR providers reported a mean monthly digital PR contract of $5,458. **In plain English:** The average monthly fee covered a complete digital PR service, not just one link. **What you can do:** 1. Ask what research, content, outreach and reporting the monthly fee includes. 2. Compare providers by the work included, not only the price. [BuzzStream: How Much Does Digital PR Cost?](https://www.buzzstream.com/blog/digital-pr-costs/) - Evidence: Survey; 2025 survey · updated page in 2026 - Sample: About 70 respondents - Population: Agency and freelance digital PR leaders, predominantly UK and US - Data period: 2025 survey; fieldwork dates not disclosed; page updated June 2026 - Source location: Average monthly contract - Limit: Self-reported service costs, not invoices. UK/US-heavy convenience sample; item-level response counts are undisclosed. Scope varies between retainers. - Evidence ID: S098; checked 2026-09-09 ### 18. Cost per link needs a definition **US$597 per link** — BuzzStream’s 2025 provider survey reported a $597 mean cost per link; roughly one-third did not know their figure. **In plain English:** The average price per link depends on what the provider included. Different definitions can produce very different totals. **What you can do:** 1. Agree what counts as a successful link before calculating cost per link. 2. Show total spend, accepted links and links that remain live together. [BuzzStream: How Much Does Digital PR Cost?](https://www.buzzstream.com/blog/digital-pr-costs/) - Evidence: Survey; 2025 survey · updated page in 2026 - Sample: About 70 respondents - Population: Agency and freelance digital PR leaders, predominantly UK and US - Data period: 2025 survey; fieldwork dates not disclosed; page updated June 2026 - Source location: Average cost per link - Limit: Self-reported service costs, not invoices. UK/US-heavy convenience sample; item-level response counts are undisclosed. Scope varies between retainers. The mean reflects respondents able to give a cost; the exact denominator is not reported. - Evidence ID: S099; checked 2026-09-09 ### 19. Provider type changes the quoted scope **US$6,357 vs US$4,200** — In that 2025 survey, mean monthly retainers were $6,357 for agencies and $4,200 for freelancers. **In plain English:** Agencies charged more on average than freelancers, but the work and capacity may not have been the same. **What you can do:** 1. Give agencies and freelancers the same brief and ask what each includes. 2. Compare the people, work and expected results beside the price. [BuzzStream: How Much Does Digital PR Cost?](https://www.buzzstream.com/blog/digital-pr-costs/) - Evidence: Survey; 2025 survey · updated page in 2026 - Sample: About 70 respondents - Population: Agency and freelance digital PR leaders, predominantly UK and US - Data period: 2025 survey; fieldwork dates not disclosed; page updated June 2026 - Source location: Agency and freelancer comparison - Limit: Self-reported service costs, not invoices. UK/US-heavy convenience sample; item-level response counts are undisclosed. Scope varies between retainers. - Evidence ID: S100; checked 2026-09-09 ### 20. Longer commitments were common **About 75% of contracts** — In BuzzStream’s 2025 provider survey, about 75% of contracts lasted longer than six months. **In plain English:** Most contracts lasted longer than six months. Buyers need review points before the full term is over. **What you can do:** 1. Add review dates after 30, 60 and 90 days to longer contracts. 2. Agree when you can change the story, audience or budget. [BuzzStream: How Much Does Digital PR Cost?](https://www.buzzstream.com/blog/digital-pr-costs/) - Evidence: Survey; 2025 survey · updated page in 2026 - Sample: About 70 respondents - Population: Agency and freelance digital PR leaders, predominantly UK and US - Data period: 2025 survey; fieldwork dates not disclosed; page updated June 2026 - Source location: Contract length - Limit: Self-reported service costs, not invoices. UK/US-heavy convenience sample; item-level response counts are undisclosed. Scope varies between retainers. - Evidence ID: S101; checked 2026-09-09 ### 21. Billable coverage can include unlinked mentions **74% / 68% / 64%** — Surveyed PR providers charged for nofollow links (74%), syndicated coverage (68%) and unlinked mentions (64%). **In plain English:** Many providers charged for mentions and copied coverage as well as standard links. Agree what counts before work starts. **What you can do:** 1. Count standard links, nofollow links, copied stories and unlinked mentions separately. 2. Agree what each result is worth before the campaign begins. [BuzzStream: How Much Does Digital PR Cost?](https://www.buzzstream.com/blog/digital-pr-costs/) - Evidence: Survey; 2025 survey · updated page in 2026 - Sample: About 70 respondents - Population: Agency and freelance digital PR leaders, predominantly UK and US - Data period: 2025 survey; fieldwork dates not disclosed; page updated June 2026 - Source location: Charging for types of coverage - Limit: Self-reported service costs, not invoices. UK/US-heavy convenience sample; item-level response counts are undisclosed. Scope varies between retainers. - Evidence ID: S102; checked 2026-09-09 ### 22. Vendor quotes differed from the lowest quote **US$461 vs US$295** — BuzzStream’s June 2026 marketplace analysis found a $461 average vendor quote versus a $295 lowest-vendor-price proxy. **In plain English:** Different sellers quoted different prices for the same publication. The cheapest number may not include the same service. **What you can do:** 1. Group duplicate publication quotes and record who offered each price. 2. Compare service, review and replacement terms before choosing the cheapest option. [BuzzStream: Cost of Guest Posts, Based on 500k Sites [Updated 2026 Data]](https://www.buzzstream.com/blog/guest-post-costs/) - Evidence: Marketplace study; Jun 2026 release · marketplace listings - Sample: 257,267 sites with price, Domain Rating and estimated traffic data - Population: Complete listings in one unnamed link marketplace - Data period: June 2026 release; collection dates undisclosed - Source location: Key takeaways; Vendor vs direct pricing - Limit: One marketplace, not the whole web. Listed quotes are not completed transactions. Traffic is an SEO-tool estimate; incomplete listings were excluded. The source calls its lowest-quote proxy “direct price”; it is not independently verified publisher-direct pricing. - Evidence ID: S103; checked 2026-09-09 ## Process and measurement Plan the work, decide what success means and measure the same things each time. ### 23. Documented processes were uncommon **32% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 32% of respondents followed a documented, repeatable link-building process. **In plain English:** Only about one in three respondents had a written process. Without one, it is harder to find where a campaign went wrong. **What you can do:** 1. Write down every step from finding websites to checking published links. 2. Give each step a simple pass or fail rule. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Link Building Strategy, question 3 - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S016; checked 2026-09-09 ### 24. Benchmarking had limited adoption **About one quarter** — In BuzzStream and Citation Labs’ 2025 survey, about one quarter used benchmarks for their link-building results. **In plain English:** Few teams compared results with a consistent starting point. Your own past results are often the fairest comparison. **What you can do:** 1. Record your usual contact, reply, coverage, link and retention rates. 2. Compare new campaigns with your own past campaigns using the same rules. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Link Building Reporting, question 12 - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S017; checked 2026-09-09 ### 25. Reports often failed to explain value **40% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 40% felt their reporting effectively communicated the value of link building. **In plain English:** Many reports listed activity without explaining business value. A live link is an output, not the final result. **What you can do:** 1. Explain which decision each number in the report should help someone make. 2. Show delivery, link quality, visits and business results separately. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Link Building Reporting, question 11 - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S018; checked 2026-09-09 ### 26. More investment was planned **56% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 56% planned to increase link-building investment over the following year. **In plain English:** More than half planned to spend more. Extra money helps only when the process can keep quality high. **What you can do:** 1. Check whether the team can keep quality high before increasing the budget. 2. Increase spending in stages and stop if quality falls. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Link Building Budgeting; next-12-month investment question - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S019; checked 2026-09-09 ### 27. Success was far from universal **29% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 29% described their link-building efforts as successful. **In plain English:** Fewer than one in three called their work successful. Teams may also mean different things when they say “success.” **What you can do:** 1. Define an acceptable, good and excellent result before the campaign begins. 2. Afterward, decide whether to stop, change the plan or do more. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Measuring Impact, question 8 - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S020; checked 2026-09-09 ### 28. Quality-link count was a common KPI **85.1% of respondents** — In BuzzStream’s 2026 digital PR survey, 85.1% measured campaign success using the number of quality links. **In plain English:** Most respondents counted high-quality links. The measure works only when quality is clearly defined. **What you can do:** 1. Write simple rules for relevance, article quality, audience, link type and retention. 2. Count only links that meet those rules and explain any rejections. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Measuring Digital PR; ways to measure success - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S023; checked 2026-09-09 ### 29. Results often required several months **85.2% said within 6 months** — In BuzzStream’s 2026 digital PR survey, 85.2% said measurable campaign results took no more than six months. **In plain English:** Most respondents said results appeared within six months. This was their experience, not a promise for every site. **What you can do:** 1. Set dates for creating the asset, outreach, first coverage and final review. 2. Tell everyone what can reasonably be judged at each date. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Measuring Digital PR; time to measurable results - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S025; checked 2026-09-09 ### 30. Digital PR felt harder **75% of respondents** — In BuzzStream’s 2026 digital PR survey, 75% found the work more challenging than a year earlier. **In plain English:** Three quarters said digital PR had become harder. Sending more emails does not fix a weak story. **What you can do:** 1. After the first emails, review the story, recipients and message separately. 2. Change one weak part, test again, then decide whether to contact more people. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Digital PR Challenges; change in difficulty - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S026; checked 2026-09-09 ### 31. Teams want better tactic selection **66% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 66% wanted to improve their choice of backlink tactics. **In plain English:** Many teams wanted better ways to choose tactics. The right tactic depends on the page, audience and story. **What you can do:** 1. Choose tactics by audience, target page and the proof you can offer. 2. Stop using tactics that repeatedly fail to produce useful results. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Areas for improvement: effective strategies - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S092; checked 2026-09-09 ### 32. Many teams align goals with KPIs **61% of respondents** — In BuzzStream and Citation Labs’ 2025 survey, 61% said their link-building goals aligned with KPIs. **In plain English:** Many teams said their goals and measurements matched. Writing both down makes performance easier to judge. **What you can do:** 1. Give each campaign goal one main success measure. 2. Do not claim success using a number that was never part of the goal. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Setting goals and KPIs - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. - Evidence ID: S094; checked 2026-09-09 ### 33. Targets often included more than followed links **62% of respondents** — In Aira’s 2022 survey, 62% counted nofollow links, including sponsored and UGC attributes, toward overall link targets. **In plain English:** Campaign targets often included nofollow links and brand mentions. These results should not be counted as the same thing. **What you can do:** 1. Set separate targets for standard links, nofollow links, mentions and copied stories. 2. Tell providers which results are required and which are optional. [Aira: 11 Takeaways from the 2022 State of Link Building Report](https://aira.net/blog/11-takeaways-2022-state-of-link-building-report/) - Evidence: Survey; Mar–May 2022 · historical survey - Sample: 270 respondents; branching questions have different denominators - Population: Self-selected SEO professionals answering the annual link-building survey - Data period: 16 March–2 May 2022 - Source location: Takeaway 4; PDF page 17, exact survey question - Limit: Historical self-reported practice and opinions. Multiple-choice questions may allow several answers; responses are not measured ranking effects. The summary also mentions brand mentions, but the PDF question specifically asks about link attributes. The selected claim follows that narrower wording. - Evidence ID: S116; checked 2026-09-09 ### 34. Weak campaigns were part of the experience **53% low-link; 29% zero-link** — In Aira’s 2022 survey, 53% had run a content campaign earning only 1–9 links, and 29% had run one earning none in the preceding year. **In plain English:** Many people had worked on campaigns that performed poorly. A useful review should explain why. **What you can do:** 1. Review weak campaigns and name the main problem: story, timing, audience or delivery. 2. Test one clear change in the next campaign. [Aira: 11 Takeaways from the 2022 State of Link Building Report](https://aira.net/blog/11-takeaways-2022-state-of-link-building-report/) - Evidence: Survey; Mar–May 2022 · historical survey - Sample: 270 respondents; branching questions have different denominators - Population: Self-selected SEO professionals answering the annual link-building survey - Data period: 16 March–2 May 2022 - Source location: Takeaway 10; content campaign results - Limit: Historical self-reported practice and opinions. Multiple-choice questions may allow several answers; responses are not measured ranking effects. Respondents could have experienced both outcomes. - Evidence ID: S118; checked 2026-09-09 ## Outreach and follow-up Reach the right people with a clear message, then track replies and coverage. ### 35. Media-list data often needed correction **62.8% of respondents** — In BuzzStream’s 2026 survey, 62.8% named outdated or incorrect journalist information as their main challenge with media-list tools. **In plain English:** Contact databases often contained wrong or old information. Check the person before you send. **What you can do:** 1. Check each important contact’s latest articles and current employer before emailing. 2. Remove or update contacts you cannot verify. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Media List Building, question 8 - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S027; checked 2026-09-09 ### 36. Most outreach emails received no reply **8.5% reply rate** — Backlinko and Pitchbox’s 2019 analysis of 12 million outreach emails found an 8.5% reply rate. **In plain English:** Only a small share of emails received a reply. Silence is common even in a large outreach campaign. **What you can do:** 1. Count delivered emails, helpful replies, coverage and accepted links separately. 2. Test the story with a small relevant group before sending it widely. [Backlinko and Pitchbox: We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) - Evidence: Website study; 2019 release · historical email dataset - Sample: 12 million outreach emails - Population: Outreach emails supplied through Pitchbox - Data period: not reported; 2019 study - Source location: Summary and most outreach emails are ignored - Limit: Historical, platform-specific observational dataset. Reply is not equivalent to a positive reply, earned link or sale. Personalization findings are associations and relative changes. - Evidence ID: S035; checked 2026-09-09 ### 37. Personalized subjects correlated with replies **30.5% higher response rate** — In Backlinko and Pitchbox’s 2019 dataset, personalized subject lines were associated with a 30.5% higher response rate. **In plain English:** Emails with personalised subjects received more replies in this study. The study cannot prove that the subject alone caused it. **What you can do:** 1. Send two clear subject lines to similar groups and compare the replies. 2. Judge useful replies and coverage, not only email opens. [Backlinko and Pitchbox: We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) - Evidence: Website study; 2019 release · historical email dataset - Sample: 12 million outreach emails - Population: Outreach emails supplied through Pitchbox - Data period: not reported; 2019 study - Source location: Personalized subject lines section - Limit: Historical, platform-specific observational dataset. Reply is not equivalent to a positive reply, earned link or sale. Personalization findings are associations and relative changes. - Evidence ID: S036; checked 2026-09-09 ### 38. Personalized bodies also correlated with replies **32.7% higher response rate** — In Backlinko and Pitchbox’s 2019 dataset, personalized email bodies were associated with a 32.7% higher response rate. **In plain English:** Personalised messages also received more replies. Useful relevance matters more than adding someone’s name. **What you can do:** 1. Explain in one sentence why the story suits that person’s recent work. 2. Compare similar email groups to see whether better relevance improves replies. [Backlinko and Pitchbox: We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) - Evidence: Website study; 2019 release · historical email dataset - Sample: 12 million outreach emails - Population: Outreach emails supplied through Pitchbox - Data period: not reported; 2019 study - Source location: Personalized email body section - Limit: Historical, platform-specific observational dataset. Reply is not equivalent to a positive reply, earned link or sale. Personalization findings are associations and relative changes. - Evidence ID: S037; checked 2026-09-09 ### 39. First coverage can arrive quickly **81% of respondents** — In BuzzStream’s 2026 digital PR survey, 81% typically saw their first coverage within a week of pitching. **In plain English:** Most surveyed teams saw their first coverage within a week. Search and business results usually take longer. **What you can do:** 1. Record how long each campaign takes to earn its first coverage. 2. Treat early coverage as feedback, not proof of final SEO or business impact. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Time to first coverage - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. Self-reported typical timing; not a campaign guarantee. - Evidence ID: S095; checked 2026-09-09 ### 40. Personalization is not universal **48% of respondents** — In BuzzStream’s 2026 digital PR survey, 48% said they always personalized pitches. **In plain English:** Fewer than half always personalised pitches. A copied message is easy for a journalist to ignore. **What you can do:** 1. Add one real, person-specific reason to every pitch. 2. Review sent emails and ignore simple name changes as personalisation. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Pitch personalization - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S096; checked 2026-09-09 ### 41. Contact checks are easy to skip **53.4% of respondents** — In BuzzStream’s 2026 digital PR survey, 53.4% always checked that a journalist still worked at the publication. **In plain English:** Only about half always checked whether the journalist still worked there. A correct email can still reach the wrong person. **What you can do:** 1. Check the person’s employer, topic area and latest article before sending. 2. Save the check date and remove old contacts. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Research before pitching - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S097; checked 2026-09-09 ### 42. One follow-up was associated with more replies **65.8% more replies** — Backlinko and Pitchbox’s 2019 analysis associated one additional follow-up with 65.8% more replies. **In plain English:** One follow-up was linked with more replies. A helpful reminder can work, while repeated pressure can backfire. **What you can do:** 1. Follow up once only when you can add useful new information. 2. Stop after a rejection, and track complaints and unsubscribe requests. [Backlinko and Pitchbox: We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) - Evidence: Website study; 2019 release · historical email dataset - Sample: 12 million outreach emails - Population: Outreach emails supplied through Pitchbox - Data period: not reported; 2019 study - Source location: Following up significantly boosts response rates - Limit: Historical, platform-specific observational dataset. Reply is not equivalent to a positive reply, earned link or sale. Personalization findings are associations and relative changes. Relative increase, not percentage points; observational comparison. - Evidence ID: S107; checked 2026-09-09 ### 43. Finding the right contact can matter **93% higher reply rate** — In the 2019 outreach dataset, contacting multiple people at an organization was associated with a 93% higher reply rate. **In plain English:** Contacting more than one relevant person was linked with more replies. It should still be done carefully. **What you can do:** 1. Choose the writer or editor whose recent work best fits the story. 2. Keep one backup and contact them only if needed. [Backlinko and Pitchbox: We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) - Evidence: Website study; 2019 release · historical email dataset - Sample: 12 million outreach emails - Population: Outreach emails supplied through Pitchbox - Data period: not reported; 2019 study - Source location: Multiple contacts at the same organization - Limit: Historical, platform-specific observational dataset. Reply is not equivalent to a positive reply, earned link or sale. Personalization findings are associations and relative changes. Relative increase; contact choice and campaign quality may confound the comparison. - Evidence ID: S108; checked 2026-09-09 ### 44. An identifiable sender may help **8.2% / 11.5% / 23.4%** — In the 2019 dataset, linked Twitter, LinkedIn and Instagram profiles were associated with 8.2%, 11.5% and 23.4% higher reply rates respectively. **In plain English:** Emails from a named person received more replies than anonymous-looking messages. Trust starts with a clear sender. **What you can do:** 1. Send from a named person with a clear signature and working reply address. 2. Test sender names while keeping the story and audience similar. [Backlinko and Pitchbox: We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) - Evidence: Website study; 2019 release · historical email dataset - Sample: 12 million outreach emails - Population: Outreach emails supplied through Pitchbox - Data period: not reported; 2019 study - Source location: Social profile links in email signatures - Limit: Historical, platform-specific observational dataset. Reply is not equivalent to a positive reply, earned link or sale. Personalization findings are associations and relative changes. Relative associations, not percentage-point increases or controlled effects. - Evidence ID: S109; checked 2026-09-09 ### 45. Question-style subjects had fewer opens **13% lower open rate** — Reboot’s undated analysis associated question-style subject lines with a 13% lower open rate. **In plain English:** Question-style subject lines were opened less often in this sample. A direct subject may be easier to understand. **What you can do:** 1. Replace vague questions with a direct description of the news or finding. 2. Compare similar groups and judge useful replies, not only opens. [Reboot Online: Digital PR Statistics: email subject-line analysis](https://www.rebootonline.com/digital-pr-statistics/) - Evidence: Website study; Undated study · one agency - Sample: More than 1,000 subject lines - Population: Subject lines from Reboot’s own outreach analysis - Data period: Collection period not reported - Source location: Email subject-line analysis - Limit: Observational agency data with limited method disclosure. Opens are not replies or placements and can be distorted by mail-client privacy features. Relative difference; campaign assignment was not randomized. - Evidence ID: S121; checked 2026-09-09 ### 46. Timely references correlated with more opens **12% higher open rate** — Reboot associated subjects containing buzzwords, such as celebrity names, events or special dates, with a 12% higher open rate. **In plain English:** Subjects tied to a recent event or article were opened more often. The reference still needs to be real and relevant. **What you can do:** 1. Mention a recent article only when it truly connects to your story. 2. Explain that connection in one short sentence. [Reboot Online: Digital PR Statistics: email subject-line analysis](https://www.rebootonline.com/digital-pr-statistics/) - Evidence: Website study; Undated study · one agency - Sample: More than 1,000 subject lines - Population: Subject lines from Reboot’s own outreach analysis - Data period: Collection period not reported - Source location: Email subject-line analysis: buzzwords - Limit: Observational agency data with limited method disclosure. Opens are not replies or placements and can be distorted by mail-client privacy features. Relative difference; the term grouping and agency sample limit generalization. - Evidence ID: S122; checked 2026-09-09 ## What journalists need Give journalists a relevant story, clear evidence and someone who can answer questions. ### 47. Some inboxes receive heavy daily volume **14% received over 20 pitches** — In Muck Rack’s January–March 2026 journalist survey, 14% said they received more than 20 PR pitches on a typical day. **In plain English:** Some journalists receive more than 20 pitches every day. They need to understand the story quickly. **What you can do:** 1. State the finding, affected audience and why it matters in the first two sentences. 2. Make the story clear without requiring an attachment. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 28, daily pitch volume - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S028; checked 2026-09-09 ### 48. Relevant pitches were scarce for many writers **43% seldom; 4% never** — In Muck Rack’s January–March 2026 survey, 43% of journalists seldom received relevant pitches, and 4% said they never did. **In plain English:** Many journalists said most pitches did not fit their work. A shorter, better-targeted list can beat a larger one. **What you can do:** 1. Read three recent articles to understand the journalist’s topics and audience. 2. Send only when your evidence clearly fits their work. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 29, pitch relevance frequency - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S029; checked 2026-09-09 ### 49. Pitches still contributed to published stories **86% of journalists** — In Muck Rack’s January–March 2026 survey, 86% of journalists said at least some of their stories began with a PR pitch. **In plain English:** Most journalists had used at least one pitch in a story. Good outreach can help when it provides useful information. **What you can do:** 1. Send a checkable finding, a short expert quote and the supporting source. 2. Record which parts journalists actually use in their coverage. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 30, stories originating from pitches - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S030; checked 2026-09-09 ### 50. Original evidence was welcome **40% of journalists** — In Muck Rack’s January–March 2026 survey, 40% of journalists wanted original data or research included with a story pitch. **In plain English:** Many journalists wanted original research. They also need enough detail to check it. **What you can do:** 1. Publish who was studied, when, how and what the limits were. 2. Offer a simple table or methods page so journalists can check the result. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 31, ideal pitch assets - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S031; checked 2026-09-09 ### 51. Audience impact made a story relevant **78% of journalists** — In Muck Rack’s January–March 2026 survey, 78% said direct effects on their audience’s community made a pitch relevant. **In plain English:** Journalists cared most about how a story affects their audience. Relevance should be obvious in the pitch. **What you can do:** 1. State who is affected, what changes and how large the change is. 2. Pitch only publications whose readers have a clear reason to care. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 33, audience relevance - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S032; checked 2026-09-09 ### 52. An irrelevant pitch was easily rejected **82% of journalists** — In Cision’s 2026 survey of more than 1,800 journalists, 82% identified irrelevance to their audience or beat as a reason to reject a pitch. **In plain English:** Irrelevant pitches were the most common reason for rejection. Sending to the wrong person wastes both sides’ time. **What you can do:** 1. Check topic fit, reader value and story format before approving a pitch. 2. Do not send when you cannot point to a relevant recent article. [Cision: Why Do Journalists Ignore Your PR Pitch?](https://www.cision.com/resources/articles/why-do-journalists-ignore-your-pr-pitch/) - Evidence: Survey; 2026 survey · 1,800+ journalists - Sample: more than 1,800 journalists across 19 markets - Population: Journalists across industries in the global State of the Media survey - Data period: not reported; 2026 survey release - Source location: 1. Make your pitch relevant - Limit: Original publisher’s public explanation of its own survey. Fieldwork dates and exact question-specific denominator not disclosed on this page; self-reported rejection reasons, not observed delivery outcomes. Full report landing page requires a form, so selected findings use accessible original-publisher reporting. - Evidence ID: S081; checked 2026-09-09 ### 53. A pitch still needs a news angle **34% of journalists** — In Cision’s 2026 journalist survey, 34% said they would reject a pitch without a clear news angle or point of view. **In plain English:** A third rejected pitches with no clear news. Data still needs a reason to matter now. **What you can do:** 1. Answer three questions: What is new? Why now? Why does it matter? 2. Improve the story before adding more contacts if the answers are weak. [Cision: Why Do Journalists Ignore Your PR Pitch?](https://www.cision.com/resources/articles/why-do-journalists-ignore-your-pr-pitch/) - Evidence: Survey; 2026 survey · 1,800+ journalists - Sample: more than 1,800 journalists across 19 markets - Population: Journalists across industries in the global State of the Media survey - Data period: not reported; 2026 survey release - Source location: 6. Bring a clear news angle or point of view - Limit: Original publisher’s public explanation of its own survey. Fieldwork dates and exact question-specific denominator not disclosed on this page; self-reported rejection reasons, not observed delivery outcomes. Full report landing page requires a form, so selected findings use accessible original-publisher reporting. - Evidence ID: S082; checked 2026-09-09 ### 54. Some journalists face a large weekly queue **29% received over 50 pitches** — In Cision’s 2026 survey, 29% of journalists received more than 50 pitches per week. **In plain English:** Some journalists receive hundreds of pitches a week. Your message must be easy to scan. **What you can do:** 1. Start with the small group of journalists who fit the story best. 2. Use their questions to improve the pitch before sending it more widely. [Cision: Why Do Journalists Ignore Your PR Pitch?](https://www.cision.com/resources/articles/why-do-journalists-ignore-your-pr-pitch/) - Evidence: Survey; 2026 survey · 1,800+ journalists - Sample: more than 1,800 journalists across 19 markets - Population: Journalists across industries in the global State of the Media survey - Data period: not reported; 2026 survey release - Source location: Pitch volume - Limit: Original publisher’s public explanation of its own survey. Fieldwork dates and exact question-specific denominator not disclosed on this page; self-reported rejection reasons, not observed delivery outcomes. Full report landing page requires a form, so selected findings use accessible original-publisher reporting. - Evidence ID: S129; checked 2026-09-09 ### 55. Most pitches missed the mark for many recipients **72% of journalists** — In Cision’s 2026 survey, 72% said fewer than a quarter of the pitches they received were relevant. **In plain English:** Many said most pitches missed the mark. Better targeting matters more than sending more. **What you can do:** 1. Sort missed pitches by wrong topic, audience, timing, format or evidence. 2. Fix repeated targeting mistakes before adding more contacts. [Cision: Why Do Journalists Ignore Your PR Pitch?](https://www.cision.com/resources/articles/why-do-journalists-ignore-your-pr-pitch/) - Evidence: Survey; 2026 survey · 1,800+ journalists - Sample: more than 1,800 journalists across 19 markets - Population: Journalists across industries in the global State of the Media survey - Data period: not reported; 2026 survey release - Source location: Pitch relevance - Limit: Original publisher’s public explanation of its own survey. Fieldwork dates and exact question-specific denominator not disclosed on this page; self-reported rejection reasons, not observed delivery outcomes. Full report landing page requires a form, so selected findings use accessible original-publisher reporting. - Evidence ID: S130; checked 2026-09-09 ### 56. A promotional angle can lose the story **53% of journalists** — In Cision’s 2026 survey, 53% would reject a pitch for being overly promotional. **In plain English:** A sales message is not automatically a news story. Lead with the useful finding. **What you can do:** 1. Lead with the useful finding or event, then explain the company’s role. 2. Remove sales claims that do not help a journalist report the story. [Cision: Why Do Journalists Ignore Your PR Pitch?](https://www.cision.com/resources/articles/why-do-journalists-ignore-your-pr-pitch/) - Evidence: Survey; 2026 survey · 1,800+ journalists - Sample: more than 1,800 journalists across 19 markets - Population: Journalists across industries in the global State of the Media survey - Data period: not reported; 2026 survey release - Source location: Reasons to reject a pitch - Limit: Original publisher’s public explanation of its own survey. Fieldwork dates and exact question-specific denominator not disclosed on this page; self-reported rejection reasons, not observed delivery outcomes. Full report landing page requires a form, so selected findings use accessible original-publisher reporting. - Evidence ID: S131; checked 2026-09-09 ### 57. PR relationships still have value **53% of journalists** — In Muck Rack’s 2026 survey, 53% rated relationships with PR professionals important or very important. **In plain English:** More than half valued relationships with PR people. Reliable help over time can make future contact easier. **What you can do:** 1. Keep short notes on each contact’s topics, preferences and past conversations. 2. Answer requests quickly and accurately, including between campaigns. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 27, printed page 26: PR relationships - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S132; checked 2026-09-09 ### 58. A non-response is common **54% of journalists** — In Muck Rack’s 2026 survey, 54% said they seldom or never replied to pitches. **In plain English:** More than half rarely or never reply. No answer does not always mean the email failed to arrive. **What you can do:** 1. Expect some silence and limit the number of follow-ups. 2. Check the story and audience fit before sending more emails. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 28: replies to pitches - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S133; checked 2026-09-09 ### 59. Access to a person helps the reporting **58% of journalists** — In Muck Rack’s 2026 survey, 58% wanted access to relevant interviewees with a pitch. **In plain English:** Many wanted access to someone they could interview. A real expert can make the evidence easier to use. **What you can do:** 1. Offer a real expert and explain what they can add. 2. Prepare them to answer questions about the method and limits. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 32, printed page 31: what to include - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S134; checked 2026-09-09 ### 60. Repeated pitches can get deleted **46% of journalists** — In Muck Rack’s 2026 survey, 46% would delete pitches sent repeatedly without a response. **In plain English:** Nearly half would delete repeated pitches. A follow-up should add something new. **What you can do:** 1. Stop after one useful follow-up unless you have genuinely new information. 2. Record the result so another teammate does not resend the same pitch. [Muck Rack: The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) - Evidence: Survey; Jan–Mar 2026 · 897 retained responses - Sample: 897 retained responses from 1,044 collected - Population: Journalists recruited by email from Muck Rack’s database; mostly US-based - Data period: 2026-01-30 to 2026-03-02 - Source location: PDF page 33: reasons to reject or delete - Limit: Non-probability, self-administered survey; not representative of every newsroom. Question-specific response counts are not always shown. Use retained sample, not the larger collection count. - Evidence ID: S135; checked 2026-09-09 ## Content that earns links Create pages that other writers genuinely want to use and cite. ### 61. Data-led pitches were widely used **95.9% of respondents** — In BuzzStream’s 2026 survey of more than 150 digital PR professionals, 95.9% used pitches built around data. **In plain English:** Almost all respondents used data-led pitches. Data helps only when it answers a question people care about. **What you can do:** 1. Choose one trustworthy number that answers a current audience question. 2. Test the story with a few relevant journalists before doing more work. [BuzzStream: State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) - Evidence: Survey; 2026 survey · fieldwork dates unstated - Sample: 150+ professionals - Population: Digital PR professionals responding to the publisher’s survey - Data period: not reported; 2026 release - Source location: Digital PR Tactics, question 5 - Limit: Self-reported vendor survey, not representative of all campaigns. Costs, success definitions and times to results differ across respondents; fieldwork dates undisclosed. - Evidence ID: S022; checked 2026-09-09 ### 62. Original research was part of many content plans **49% of respondents** — In Orbit Media’s August 2025 survey of 808 content marketers, 49% had created and published original research in the preceding year. **In plain English:** About half used original research. First-party evidence can give other writers something new to cite. **What you can do:** 1. List useful company data, customer questions and repeated observations. 2. Check privacy, definitions and whether you can update the study later. [Orbit Media: Annual Blogger Survey 2025](https://www.orbitmedia.com/blog/blogging-statistics/) - Evidence: Survey; Aug 2025 · 808 content marketers - Sample: 808 respondents - Population: Individual bloggers/content marketers; sample skews toward US, B2B and the author’s LinkedIn network - Data period: August 2025 - Source location: Question 20 chart, original research adoption; methodology - Limit: Voluntary survey with no incentives. Strong results are self-defined, not a uniform SEO or revenue metric. Cross-sectional associations are not causal. Selected chart values visually inspected. - Evidence ID: S049; checked 2026-09-09 ### 63. Statistics were a common content ingredient **60% of respondents** — In Orbit Media’s August 2025 survey, 60% said their typical blog post included statistics. **In plain English:** Most used statistics in their content. Numbers are more useful when readers can trace them to the source. **What you can do:** 1. Put the source, date and people studied beside every important number. 2. Give each finding a permanent link so writers can cite it easily. [Orbit Media: Annual Blogger Survey 2025](https://www.orbitmedia.com/blog/blogging-statistics/) - Evidence: Survey; Aug 2025 · 808 content marketers - Sample: 808 respondents - Population: Individual bloggers/content marketers; sample skews toward US, B2B and the author’s LinkedIn network - Data period: August 2025 - Source location: Question 10 chart, elements in typical posts - Limit: Voluntary survey with no incentives. Strong results are self-defined, not a uniform SEO or revenue metric. Cross-sectional associations are not causal. Selected chart values visually inspected. - Evidence ID: S050; checked 2026-09-09 ### 64. Guest bloggers reported stronger outcomes **30% vs 15% strong results** — In Orbit Media’s August 2025 survey, 30% of guest bloggers reported strong results, compared with 15% of those who did not guest blog. **In plain English:** Guest bloggers were more likely to report strong results. The survey does not show that the link caused those results. **What you can do:** 1. Choose guest articles for the right audience and editorial fit. 2. Measure visits, links and relationships separately, and avoid claiming one article changed rankings. [Orbit Media: Annual Blogger Survey 2025](https://www.orbitmedia.com/blog/blogging-statistics/) - Evidence: Survey; Aug 2025 · 808 content marketers - Sample: 808 respondents - Population: Individual bloggers/content marketers; sample skews toward US, B2B and the author’s LinkedIn network - Data period: August 2025 - Source location: Question 17 chart, guest blogging and results - Limit: Voluntary survey with no incentives. Strong results are self-defined, not a uniform SEO or revenue metric. Cross-sectional associations are not causal. Selected chart values visually inspected. - Evidence ID: S051; checked 2026-09-09 ### 65. Teams using analytics reported better results **32% vs 13% reported strong results** — In Orbit Media’s August 2025 survey, 32% of bloggers who always used analytics reported strong results, versus 13% who rarely or never did. **In plain English:** People who always checked their analytics were more likely to say their content worked well. The survey does not prove why. **What you can do:** 1. Choose one clear goal for each article, such as visits, leads or links, and check it every month. 2. Write down what you changed so you can see which updates were followed by better results. [Orbit Media: Annual Blogger Survey 2025](https://www.orbitmedia.com/blog/blogging-statistics/) - Evidence: Survey; Aug 2025 · 808 content marketers - Sample: 808 respondents - Population: Individual bloggers/content marketers; sample skews toward US, B2B and the author’s LinkedIn network - Data period: August 2025 - Source location: Question 21 chart, analytics and results - Limit: Voluntary survey with no incentives. Strong results are self-defined, not a uniform SEO or revenue metric. Cross-sectional associations are not causal. Selected chart values visually inspected. - Evidence ID: S052; checked 2026-09-09 ### 66. Most AI users still checked the work **80% of respondents** — In Ahrefs’ December 2024–February 2025 survey, 80% of respondents reported manually reviewing AI content for accuracy. **In plain English:** Most people in this survey reviewed AI-assisted content themselves. Human checking matters when an article needs to earn trust. **What you can do:** 1. Ask a person to check every important fact, name and source before AI-assisted content is published. 2. Link to the original evidence and correct mistakes quickly when a reader finds one. [Ahrefs: Marketers Using AI Publish 42% More Content](https://ahrefs.com/blog/marketers-using-ai-publish-more-content/) - Evidence: Survey; Dec 2024–Feb 2025 · marketer survey - Sample: 879 reported total responses - Population: Marketing and SEO professionals responding to an open content-marketing survey - Data period: December 2024–February 2025 - Source location: Companies edit and review AI content - Limit: Open, self-selected survey; US is the most common location, in-house the most common role type. Question-specific denominators vary. Report says 879 total but its printed AI/non-AI subgroup counts sum to 880; use reported rounded findings without recomputing subgroup shares. Associations cannot isolate an AI effect. - Evidence ID: S055; checked 2026-09-09 ### 67. Misinformation was the main AI concern **62% of respondents** — In Ahrefs’ December 2024–February 2025 survey, 62% identified sharing misinformation as the biggest perceived risk of AI content. **In plain English:** Incorrect information was the most common concern about AI-assisted content. People will not cite a source they cannot trust. **What you can do:** 1. Trace every number and quote back to its original source before using it in a linkable article. 2. Remove a claim when you cannot confirm it instead of repeating a confident-sounding AI answer. [Ahrefs: Websites Using AI Content Grow 5% Faster](https://ahrefs.com/blog/websites-using-ai-content-grow-faster/) - Evidence: Survey; Dec 2024–Feb 2025 · marketer survey - Sample: 879 reported total responses - Population: Marketing and SEO professionals in Ahrefs’ AI content-marketing survey - Data period: December 2024–February 2025 - Source location: What are the perceived risks of AI content? - Limit: Open, self-selected survey; US is the most common location, in-house the most common role type. Question-specific denominators vary. Report says 879 total but its printed AI/non-AI subgroup counts sum to 880; use reported rounded findings without recomputing subgroup shares. Associations cannot isolate an AI effect. Selected finding concerns perceived risk, not actual misinformation incidence; headline growth calculation is not selected. - Evidence ID: S058; checked 2026-09-09 ### 68. Blogs and product pages play different roles **68% vs 41%** — In BuzzStream and Citation Labs’ 2025 survey, 68% prioritized blog posts for links and 41% prioritized product pages. **In plain English:** Teams often built links to blog posts more than product pages. A useful guide can lead readers toward a product later. **What you can do:** 1. Link each useful guide or study to the product page it supports. 2. Make the next step clear without turning the guide into a sales page. [BuzzStream and Citation Labs: Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) - Evidence: Survey; 2025 survey · fieldwork dates unstated - Sample: Not disclosed - Population: Agency, freelance, consultancy and in-house link-building professionals - Data period: not reported; 2025 release - Source location: Types of pages prioritized for backlinks - Limit: Self-reported, non-probability survey; full sample size not stated. The approximately 150 respondents cited in the management section are the in-house subset only. Multiple page types could be selected. - Evidence ID: S093; checked 2026-09-09 ### 69. Longer articles attracted more referring domains **77.2% more linking domains** — In the 2017–2018 content sample, articles over 3,000 words averaged 77.2% more referring domains than articles below 1,000 words. **In plain English:** Longer articles earned more links on average. Extra words alone are unlikely to be the reason. **What you can do:** 1. Compare your guide with the best sources and add missing evidence or examples. 2. Remove filler; make the article longer only when readers need more detail. [Backlinko / BuzzSumo: We Analyzed 912 Million Blog Posts. Here’s What We Learned About Content Marketing](https://backlinko.com/content-study) - Evidence: Website study; 2017–2018 articles · 2019 release - Sample: 912.5 million articles - Population: Articles in the BuzzSumo index - Data period: Articles published 1 October 2017–1 October 2018 - Source location: Long-form content gets more backlinks; methods PDF - Limit: Historical, observational data. Links are unique referring domains in the methods. Means can be skewed by outliers; formats were classified automatically. - Evidence ID: S110; checked 2026-09-09 ### 70. Most articles attracted no external links **94% had no linking domains** — In the historical Backlinko/BuzzSumo sample, 94% of articles had no external referring domains; only 2.2% had links from multiple sites. **In plain English:** Most articles earned no links from other websites. Good content still needs a useful topic and active promotion. **What you can do:** 1. Before writing, find who links to similar resources and why. 2. Plan promotion, internal links and future updates before publishing. [Backlinko / BuzzSumo: We Analyzed 912 Million Blog Posts. Here’s What We Learned About Content Marketing](https://backlinko.com/content-study) - Evidence: Website study; 2017–2018 articles · 2019 release - Sample: 912.5 million articles - Population: Articles in the BuzzSumo index - Data period: Articles published 1 October 2017–1 October 2018 - Source location: 94% of blog posts have zero external links - Limit: Historical, observational data. Links are unique referring domains in the methods. Means can be skewed by outliers; formats were classified automatically. - Evidence ID: S111; checked 2026-09-09 ### 71. Social shares were a weak link proxy **0.078 correlation** — Backlinko and BuzzSumo reported a 0.078 Pearson correlation between social shares and backlinks in their historical content dataset. **In plain English:** Social shares and links did not move together closely. A popular post is not always a cited post. **What you can do:** 1. Report social shares, visits and new linking websites separately. 2. Use each number for its own decision; shares do not prove links will follow. [Backlinko / BuzzSumo: We Analyzed 912 Million Blog Posts. Here’s What We Learned About Content Marketing](https://backlinko.com/content-study) - Evidence: Website study; 2017–2018 articles · 2019 release - Sample: 912.5 million articles - Population: Articles in the BuzzSumo index - Data period: Articles published 1 October 2017–1 October 2018 - Source location: No correlation between social shares and backlinks - Limit: Historical, observational data. Links are unique referring domains in the methods. Means can be skewed by outliers; formats were classified automatically. - Evidence ID: S112; checked 2026-09-09 ### 72. Some explanatory formats attracted more links **25.8% more linking domains** — Why posts, what posts and infographics averaged 25.8% more links than video and how-to posts in that historical sample. **In plain English:** Why-posts, how-to guides and infographics earned more links on average. Choose the format that fits the subject. **What you can do:** 1. Choose a why-post, how-to guide or visual only when it fits the question. 2. Compare it with similar content and keep promotion differences in mind. [Backlinko / BuzzSumo: We Analyzed 912 Million Blog Posts. Here’s What We Learned About Content Marketing](https://backlinko.com/content-study) - Evidence: Website study; 2017–2018 articles · 2019 release - Sample: 912.5 million articles - Population: Articles in the BuzzSumo index - Data period: Articles published 1 October 2017–1 October 2018 - Source location: Content formats and backlinks - Limit: Historical, observational data. Links are unique referring domains in the methods. Means can be skewed by outliers; formats were classified automatically. Grouped categories; this does not estimate the effect of changing the format of one article. - Evidence ID: S113; checked 2026-09-09 ### 73. A statistics page earned links in one campaign **515 emails → 32 linking sites** — Ahrefs’ 2020 statistics-page campaign reported 36 backlinks from 32 websites after 515 outreach emails. **In plain English:** One Ahrefs statistics page earned links through outreach. It is a useful example, not a guaranteed result. **What you can do:** 1. Give each statistic a permanent link, visible source and clear date. 2. When you update it, tell people who already cited the page what changed. [Ahrefs: Link Building Case Study: How We Built Backlinks With a Stats Page](https://ahrefs.com/blog/link-building-case-study/) - Evidence: One campaign; 2020 case study · one campaign - Sample: 515 emails - Population: One Ahrefs statistics-page outreach campaign - Data period: Campaign reported August 2020; precise outreach window undisclosed - Source location: Campaign results in the introduction - Limit: One successful, publisher-selected case with an established brand. No control group; the results are not a general outreach forecast. - Evidence ID: S114; checked 2026-09-09 ### 74. Reports were rated ahead of interactives **56% vs 42%** — In Aira’s 2022 survey, 56% considered reports and long-form content effective, versus 42% for interactive content. **In plain English:** Respondents rated research reports highly as linkable assets. A report works when its evidence is genuinely useful. **What you can do:** 1. Start with a report when original evidence is the main value. 2. Add clear charts, tables and summaries that publishers can reuse accurately. [Aira: 11 Takeaways from the 2022 State of Link Building Report](https://aira.net/blog/11-takeaways-2022-state-of-link-building-report/) - Evidence: Survey; Mar–May 2022 · historical survey - Sample: 270 respondents; branching questions have different denominators - Population: Self-selected SEO professionals answering the annual link-building survey - Data period: 16 March–2 May 2022 - Source location: Takeaway 11; PDF page 47: effective content formats - Limit: Historical self-reported practice and opinions. Multiple-choice questions may allow several answers; responses are not measured ranking effects. Practitioner opinions about effectiveness, not measured link yields. - Evidence ID: S119; checked 2026-09-09 ### 75. Link earning per visit varied by industry **14.4 vs 2.5 domains** — Siege’s 2024–2025 real estate/property/home-services group averaged 14.4 new referring domains per 10,000 organic visits, versus 2.5 for B2B SaaS. **In plain English:** The number of new linking websites per visit varied a lot by industry. Compare your result with similar businesses. **What you can do:** 1. Compare new linking websites with search visits using the same dates and tool. 2. Compare only with similar sites, and check both numbers before drawing conclusions. [Siege Media: Organic Link Ratios by Industry: What 100 Sites Reveal About Link Acquisition](https://www.siegemedia.com/research/organic-link-ratios) - Evidence: Website study; 2024–2025 data · 2026 release - Sample: 100 sites; ten per industry - Population: Sites selected for organic prominence across ten industries - Data period: January 2024–December 2025 - Source location: Industry comparison table; two-year averages - Limit: Organic link ratio means new referring domains per 10,000 organic visits. Selected successful sites are not a random sample. Falling traffic can raise the ratio without an increase in referring domains. - Evidence ID: S123; checked 2026-09-09 ### 76. Zillow’s link-to-traffic ratio increased **31.9 → 53.9** — Zillow’s reported organic link ratio rose from 31.9 in 2024 to 53.9 in 2025 in Siege’s analysis. **In plain English:** Zillow’s ratio rose, but that can happen because links grow, traffic falls, or both. **What you can do:** 1. Show new linking websites and search visits as separate lines. 2. Find which pages earned links before copying a single company’s result. [Siege Media: Organic Link Ratios by Industry: What 100 Sites Reveal About Link Acquisition](https://www.siegemedia.com/research/organic-link-ratios) - Evidence: Website study; 2024–2025 data · 2026 release - Sample: One website within a 100-site study - Population: Zillow in the 2024–2025 study - Data period: January 2024–December 2025 - Source location: Real estate: Zillow - Limit: Organic link ratio means new referring domains per 10,000 organic visits. Selected successful sites are not a random sample. Falling traffic can raise the ratio without an increase in referring domains. Units: new referring domains per 10,000 organic visits; not a raw backlink growth rate. - Evidence ID: S124; checked 2026-09-09 ### 77. Secureframe’s ratio also increased **3.9 → 20.1** — Secureframe’s reported organic link ratio increased from 3.9 in 2024 to 20.1 in 2025. **In plain English:** Secureframe’s ratio also rose. A single company can show a pattern, not a universal benchmark. **What you can do:** 1. Find groups of content whose link-to-visit ratio is rising. 2. Use the pattern as an idea to test on another group. [Siege Media: Organic Link Ratios by Industry: What 100 Sites Reveal About Link Acquisition](https://www.siegemedia.com/research/organic-link-ratios) - Evidence: Website study; 2024–2025 data · 2026 release - Sample: One website within a 100-site study - Population: Secureframe in the 2024–2025 study - Data period: January 2024–December 2025 - Source location: Cybersecurity: Secureframe - Limit: Organic link ratio means new referring domains per 10,000 organic visits. Selected successful sites are not a random sample. Falling traffic can raise the ratio without an increase in referring domains. Units: new referring domains per 10,000 organic visits; not an isolated campaign experiment. - Evidence ID: S125; checked 2026-09-09 ### 78. The ratio can fall sharply too **19.9 → 1.8** — WooCommerce’s reported organic link ratio fell from 19.9 in 2024 to 1.8 in 2025. **In plain English:** WooCommerce’s ratio fell sharply. Ratios can hide what happened to links and traffic separately. **What you can do:** 1. When the ratio changes sharply, check links and visits separately. 2. Show both original numbers beside the ratio in every report. [Siege Media: Organic Link Ratios by Industry: What 100 Sites Reveal About Link Acquisition](https://www.siegemedia.com/research/organic-link-ratios) - Evidence: Website study; 2024–2025 data · 2026 release - Sample: One website within a 100-site study - Population: WooCommerce in the 2024–2025 study - Data period: January 2024–December 2025 - Source location: Ecommerce: WooCommerce - Limit: Organic link ratio means new referring domains per 10,000 organic visits. Selected successful sites are not a random sample. Falling traffic can raise the ratio without an increase in referring domains. Units: new referring domains per 10,000 organic visits; the ratio alone cannot identify the cause. - Evidence ID: S126; checked 2026-09-09 ## Keeping links useful Keep checking links and pages after a campaign ends. ### 79. Links disappear over time **66.5% cumulative decay** — In Ahrefs’ historical study of links to 2,062,173 websites, 66.5% of links tracked from January 2013 had decayed by the study endpoint. **In plain English:** Many links disappeared over nine years. A link is not permanent unless someone keeps checking it. **What you can do:** 1. Check acquired links after 30, 90 and 180 days. 2. Record why a link disappeared and try to recover only fixable losses. [Ahrefs: At Least 66.5% of Links to Sites in the Last 9 Years Are Dead](https://ahrefs.com/blog/link-rot-study/) - Evidence: Website study; From Jan 2013 · about nine years - Sample: 2,062,173 websites - Population: Historical backlinks pointing to sampled websites - Data period: Links tracked from January 2013 across a described nine-year window; exact endpoint not reported - Source location: Opening finding and link-rot methodology - Limit: Page updated February 2, 2024; do not substitute that date for measurement or original publication. Cumulative decay is not an annual loss rate. Ahrefs crawler classifications determine status. - Evidence ID: S038; checked 2026-09-09 ### 80. Original page addresses often stopped working **25% of sampled pages** — Pew found that 25% of sampled pages that existed at some point during 2013–2023 were inaccessible at their original addresses in October 2023. **In plain English:** One quarter of old pages no longer worked at their original address. Sources and destinations can both move. **What you can do:** 1. Save both the page containing the link and the page it points to. 2. Redirect moved pages and update your campaign records when addresses change. [Pew Research Center: When Online Content Disappears](https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears/) - Evidence: Website study; 2023 observations · 2024 release - Sample: approximately 1 million historical pages; about 500,000 news and 500,000 government pages; 50,000 Wikipedia pages - Population: Common Crawl pages; sampled news, government and English-language Wikipedia pages - Data period: 2013–2023 Common Crawl pages checked October 2023; news, government and Wikipedia samples from spring 2023 - Source location: A quarter of all webpages that existed at one point between 2013 and 2023 are no longer accessible - Limit: Availability at the original address was checked at a particular time. Pages and links have different denominators; a moved page may count as inaccessible at its former address. - Evidence ID: S039; checked 2026-09-09 ### 81. Broken links appeared on established publishers **23% news; 21% government pages** — In Pew’s spring 2023 samples, 23% of news pages and 21% of government pages contained at least one broken link. **In plain English:** Broken links appeared even on trusted news and government sites. Famous domains still have page-level problems. **What you can do:** 1. Check that the page works, appears in Google and still contains the link. 2. Check again after delivery, even on famous websites. [Pew Research Center: When Online Content Disappears](https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears/) - Evidence: Website study; 2023 observations · 2024 release - Sample: approximately 1 million historical pages; about 500,000 news and 500,000 government pages; 50,000 Wikipedia pages - Population: Common Crawl pages; sampled news, government and English-language Wikipedia pages - Data period: 2013–2023 Common Crawl pages checked October 2023; news, government and Wikipedia samples from spring 2023 - Source location: Broken links are common on government and news websites - Limit: Availability at the original address was checked at a particular time. Pages and links have different denominators; a moved page may count as inaccessible at its former address. - Evidence ID: S040; checked 2026-09-09 ### 82. Reference lists also deteriorated **11% of Wikipedia references** — In Pew’s spring 2023 sample of 50,000 English Wikipedia pages, 11% of reference links were inaccessible. **In plain English:** Some Wikipedia references also broke. Research pages need regular source checks. **What you can do:** 1. Check links on your research pages regularly, starting with popular pages. 2. Replace dead sources and record what changed and when. [Pew Research Center: When Online Content Disappears](https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears/) - Evidence: Website study; 2023 observations · 2024 release - Sample: approximately 1 million historical pages; about 500,000 news and 500,000 government pages; 50,000 Wikipedia pages - Population: Common Crawl pages; sampled news, government and English-language Wikipedia pages - Data period: 2013–2023 Common Crawl pages checked October 2023; news, government and Wikipedia samples from spring 2023 - Source location: Links on Wikipedia; reference-link analysis - Limit: Availability at the original address was checked at a particular time. Pages and links have different denominators; a moved page may count as inaccessible at its former address. - Evidence ID: S041; checked 2026-09-09 ## Internal links and site structure Help people and search engines find the important pages on your own site. ### 83. Many domains had sparsely linked pages **66.2% of domains** — In Ahrefs’ 2023 audit sample, 66.2% of domains had at least one page with only one followed incoming internal link. **In plain English:** Many sites had pages with few internal links. Those pages can be harder for people and search engines to find. **What you can do:** 1. Find important pages that receive few links from other pages on your site. 2. Add links from related pages using clear words that describe the destination. [Ahrefs: We Crawled Over 1 Million Domains to Find the Most Common Technical SEO Issues](https://ahrefs.com/blog/site-audit-study/) - Evidence: Website study; 2023 release · audited domains - Sample: more than 1 million domains - Population: Domains processed through Ahrefs Site Audit; a domain counts if at least one page has the issue - Data period: not reported; August 2023 report - Source location: Most common issues; incoming internal links - Limit: Site Audit sample is not a random sample of the entire web. Percentages apply to domains with at least one affected page, not the proportion of all pages. - Evidence ID: S042; checked 2026-09-09 ### 84. Redirect links were common in crawls **62.7% of domains** — In Ahrefs’ 2023 audit sample, 62.7% of domains had at least one page linking to a redirect. **In plain English:** Many sites linked through a redirect instead of the final page. Updating the link can make the journey cleaner. **What you can do:** 1. Find internal links that first pass through a redirect. 2. Point important links straight to the final page while keeping useful external redirects. [Ahrefs: We Crawled Over 1 Million Domains to Find the Most Common Technical SEO Issues](https://ahrefs.com/blog/site-audit-study/) - Evidence: Website study; 2023 release · audited domains - Sample: more than 1 million domains - Population: Domains processed through Ahrefs Site Audit; a domain counts if at least one page has the issue - Data period: not reported; August 2023 report - Source location: Most common issues; links to redirects - Limit: Site Audit sample is not a random sample of the entire web. Percentages apply to domains with at least one affected page, not the proportion of all pages. - Evidence ID: S043; checked 2026-09-09 ### 85. More internal links correlated with more clicks **4× average Google clicks** — In Zyppy’s undated internal-link study, URLs with 40–44 incoming internal links averaged four times the Google clicks of URLs with zero to four. **In plain English:** Pages with more internal links also got more Google clicks. More important pages may naturally have both. **What you can do:** 1. Compare similar pages by internal links, search clicks and purpose. 2. Test one navigation change on a small group before changing the whole site. [Zyppy: 23 Million Internal Links – SEO Case Study](https://zyppy.com/seo/seo-study/) - Evidence: Website study; Undated study · Google click data - Sample: 23 million internal links, 1,800 websites, approximately 520,000 URLs - Population: URLs on participating websites, with internal-link counts compared to Google Search Console clicks - Data period: not reported - Source location: More internal links correlate with higher traffic - Limit: Study and measurement dates undisclosed; page updated February 23, 2026. Observational averages can reflect page importance and other confounders. Results do not identify an optimal link count or causal anchor-text effect. - Evidence ID: S045; checked 2026-09-09 ### 86. Descriptive anchors had a strong association **At least 5× search traffic** — In Zyppy’s undated study, pages with at least one exact-match internal anchor received at least five times the search traffic of pages without one. **In plain English:** Pages with clear link text received more traffic on average. The study cannot prove that the wording caused the difference. **What you can do:** 1. Find vague link text such as “click here” on important user journeys. 2. Rewrite it to describe the destination in a natural way. [Zyppy: 23 Million Internal Links – SEO Case Study](https://zyppy.com/seo/seo-study/) - Evidence: Website study; Undated study · Google click data - Sample: 23 million internal links, 1,800 websites, approximately 520,000 URLs - Population: URLs on participating websites, with internal-link counts compared to Google Search Console clicks - Data period: not reported - Source location: Exact Match Anchors and Google Search Traffic - Limit: Study and measurement dates undisclosed; page updated February 23, 2026. Observational averages can reflect page importance and other confounders. Results do not identify an optimal link count or causal anchor-text effect. - Evidence ID: S046; checked 2026-09-09 ### 87. A category-link test improved traffic **+25% estimated traffic** — SearchPilot’s August 2020 grocery-site case study estimated a 25% organic-traffic increase across the tested category pages after adding links between category levels. **In plain English:** One website gained traffic after connecting related category pages. Your site may respond differently. **What you can do:** 1. Connect one group of related category pages where navigation is weak. 2. Compare visits and user behaviour with a similar unchanged group. [SearchPilot: SEO Split Test Lessons: Increasing Internal Linking](https://www.searchpilot.com/resources/case-studies/seo-split-test-lessons-increasing-internal-linking) - Evidence: SEO test; 2020 case study · one website - Sample: Not disclosed - Population: Category pages on one grocery website; links added between category levels - Data period: not reported; case study published August 2020 - Source location: Results of increasing internal linking experiment - Limit: One client experiment using page groups and a modeled control/counterfactual. The estimated effect does not establish an expected uplift for other sites; page count and experiment dates not reported. - Evidence ID: S047; checked 2026-09-09 ### 88. A positive estimate can remain uncertain **+10.2%; inconclusive at 95%** — SearchPilot’s April 2024 link-block experiment estimated a 10.2% organic-traffic increase, but the result was inconclusive at 95% confidence. **In plain English:** The test showed a possible gain, but the evidence was not strong enough to call it a clear win. **What you can do:** 1. Decide in advance how large and reliable a test result must be. 2. Record unclear results as learning, not proof that the change worked. [SearchPilot: SEO Split Test Lessons: Reducing a Link Block](https://www.searchpilot.com/resources/case-studies/seo-split-test-lessons-reducing-link-block) - Evidence: SEO test; 2024 case study · one website - Sample: Not disclosed - Population: One website’s desktop internal-link block, shortened with a Show More control - Data period: not reported; case study published April 2024 - Source location: Results of reducing link block experiment - Limit: Estimated result is inconclusive at 95% confidence, although positive at 90%. Do not describe it as a demonstrated gain or general advice to hide links; test sample not reported. - Evidence ID: S048; checked 2026-09-09 ### 89. Empty links appeared on nearly half of homepages **46.3% of homepages** — WebAIM’s February 2026 scan found empty links on 46.3% of the tested homepages. **In plain English:** An empty link gives people little or no clue about where it goes. It can also make the page harder to use. **What you can do:** 1. Ask your developer or site-audit tool to find links that have no words or useful image label. 2. Give each link a clear name that explains where it goes, or remove it when it serves no purpose. [WebAIM: The WebAIM Million: The 2026 Report on the Accessibility of the Top 1,000,000 Home Pages](https://webaim.org/projects/million/) - Evidence: Website study; Feb 2026 · one million homepages - Sample: 1 million homepages - Population: Top homepages selected using the Tranco ranking and evaluated with WAVE - Data period: February 2026 - Source location: Most common errors; empty links - Limit: Automated detection of WCAG-related failures; not a complete accessibility evaluation or legal conformance judgment. Sample is homepages, not all web pages. - Evidence ID: S066; checked 2026-09-09 ### 90. Vague link text appeared on many homepages **15.2% of homepages** — WebAIM’s February 2026 scan found ambiguous link text, such as “click here,” on 15.2% of tested homepages. **In plain English:** Text such as “click here” does not explain the destination. Clear wording helps people decide whether a link is useful. **What you can do:** 1. Replace vague phrases such as “click here” with words that describe the page people will open. 2. Read the link text on its own; if the destination is unclear, rewrite it. [WebAIM: The WebAIM Million: The 2026 Report on the Accessibility of the Top 1,000,000 Home Pages](https://webaim.org/projects/million/) - Evidence: Website study; Feb 2026 · one million homepages - Sample: 1 million homepages - Population: Top homepages selected using the Tranco ranking and evaluated with WAVE - Data period: February 2026 - Source location: Links; ambiguous link text - Limit: Automated detection of WCAG-related failures; not a complete accessibility evaluation or legal conformance judgment. Sample is homepages, not all web pages. - Evidence ID: S067; checked 2026-09-09 ### 91. Internal links also show an association **0.117 correlation** — Incoming internal links had a 0.117 ranking correlation in the same Ahrefs sample. **In plain English:** Pages with more incoming internal links tended to rank a little better. This was a relationship, not proof. **What you can do:** 1. Ask a site-audit tool to find important pages with few internal links. 2. Add links from related strong pages, then watch whether discovery and search visits improve. [Ahrefs: Google Says “Links Matter Less”—We Looked at 1,000,000 SERPs to See if It’s True](https://ahrefs.com/blog/links-matter-less-but-still-matter/) - Evidence: Website study; 2025 release · observation dates unstated - Sample: 1,000,000 keywords - Population: Top 20 Google results for the one million highest-volume US keywords - Data period: not reported; published January 2025 - Source location: Correlation table: Internal links - Limit: Observational Spearman correlations; high-volume US queries are not representative of all searches. Correlation cannot establish ranking causation. - Evidence ID: S084; checked 2026-09-09 ### 92. URL anchors had a traffic association **Almost 50% more traffic** — In Zyppy’s undated study, pages with naked-URL internal anchors had almost 50% more traffic than pages without them. **In plain English:** Pages with URL-shaped link text got more traffic on average. That does not mean raw URLs are better for readers. **What you can do:** 1. Find internal links that use the full web address as their text. 2. Keep them only when the address helps readers; otherwise use descriptive words. [Zyppy: 23 Million Internal Links – SEO Case Study](https://zyppy.com/seo/seo-study/) - Evidence: Website study; Undated study · Google click data - Sample: 23 million internal links, 1,800 websites, approximately 520,000 URLs - Population: URLs on participating websites, with internal-link counts compared to Google Search Console clicks - Data period: not reported - Source location: Naked URL anchors - Limit: Study and measurement dates undisclosed; page updated February 23, 2026. Observational averages can reflect page importance and other confounders. Results do not identify an optimal link count or causal anchor-text effect. - Evidence ID: S087; checked 2026-09-09 ### 93. Some internal links had no anchor text **Over 6% of links** — More than 6% of internal links in Zyppy’s sample had no anchor text. **In plain English:** Some internal links had no words. Many may have been image links, so each case needs checking. **What you can do:** 1. Find links with no text and check whether linked images have useful labels. 2. Fix broken or invisible controls first, and keep legitimate image links. [Zyppy: 23 Million Internal Links – SEO Case Study](https://zyppy.com/seo/seo-study/) - Evidence: Website study; Undated study · Google click data - Sample: 23 million internal links, 1,800 websites, approximately 520,000 URLs - Population: URLs on participating websites, with internal-link counts compared to Google Search Console clicks - Data period: not reported - Source location: Anchor text analysis - Limit: Study and measurement dates undisclosed; page updated February 23, 2026. Observational averages can reflect page importance and other confounders. Results do not identify an optimal link count or causal anchor-text effect. The total study covered about 23 million internal links. Missing text can include image links; it is not automatically an SEO penalty. - Evidence ID: S088; checked 2026-09-09 ### 94. A footer-link test differed by device **+10% desktop; +5% mobile** — SearchPilot’s 2021 footer-link experiment estimated a 10% desktop traffic lift at 95% confidence and 5% mobile lift at 90% confidence. **In plain English:** A footer-link test helped desktop pages more clearly than mobile pages. Results can differ by device. **What you can do:** 1. Compare the test results for desktop and mobile separately. 2. Expand the change only when it works well and remains easy to use. [SearchPilot: Does adding internal links to your home page’s footer improve organic traffic?](https://www.searchpilot.com/resources/case-studies/adding-internal-links-to-home-page-footer) - Evidence: SEO test; 2021 case study · one website - Sample: One controlled page-split experiment; page count undisclosed - Population: Linked-to pages on one customer site after expanding homepage footer links - Data period: Test reported September 2021; exact dates not reported - Source location: Device-level results after expanding roughly 30 footer links to 100+ - Limit: Desktop lift met 95% confidence; mobile met 90%, not 95%. Combined estimate was inconclusive at 95%. One customer and one implementation. - Evidence ID: S127; checked 2026-09-09 ### 95. A breadcrumb change reduced traffic **−5.5% organic traffic** — A SearchPilot 2024 ecommerce test reported a statistically significant 5.5% traffic decline after removing self-referential breadcrumb links. **In plain English:** Traffic fell when one site removed a breadcrumb link. The test does not prove that every self-link should stay. **What you can do:** 1. Review visible breadcrumbs, internal links and breadcrumb code together. 2. Test on a small group before changing the whole website. [SearchPilot: Will removing self-referential breadcrumb links improve organic traffic?](https://www.searchpilot.com/resources/case-studies/removing-self-referential-breadcrumb) - Evidence: SEO test; 2024 case study · one website - Sample: One controlled page-split experiment; page count undisclosed - Population: Product listing pages on one ecommerce customer site - Data period: Test reported July 2024; exact dates not reported - Source location: Results; implementation and schema mismatch - Limit: The visible breadcrumb changed while its schema did not. The author’s proposed explanation is not separately tested; this cannot isolate a universal effect of self-links. - Evidence ID: S128; checked 2026-09-09 ## AI citations and visibility Track AI mentions, citations and website visits as separate results. ### 96. AI summaries coincided with fewer result clicks **8% vs 15% of visits** — In Pew’s March 2025 US browsing study, traditional-result clicks occurred on 8% of Google visits with an AI summary, versus 15% without one. **In plain English:** People clicked normal Google results less often when an AI summary appeared. The searches were not identical, so treat this as a pattern. **What you can do:** 1. Track important searches with and without an AI summary. 2. Compare similar searches before blaming the summary for fewer clicks. [Pew Research Center: Google Users Are Less Likely to Click on Links When an AI Summary Appears in the Results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) - Evidence: Website study; Mar 2025 browsing · US adults - Sample: 900 adults; 68,879 unique searches reconstructed - Population: Google search visits in browsing histories of US adults in a tracking panel - Data period: Browsing March 1–31, 2025; search-result pages reconstructed April 7–17, 2025 - Source location: Clicking on traditional search results - Limit: Observational comparison, with different query types in AI and non-AI groups. April reconstructions may differ from March results; US panel only. A browsing session ends when the user exits the browser for at least five seconds. - Evidence ID: S068; checked 2026-09-09 ### 97. Clicks on summary citations were uncommon **1% of visits with a summary** — In Pew’s March 2025 US browsing study, a link within the AI summary was clicked on 1% of visits with a summary. **In plain English:** People rarely clicked links inside AI summaries. Being cited and receiving a visit are different results. **What you can do:** 1. Record AI citations and visits from AI tools as separate results. 2. Report a citation as visibility, even when it produces no visit. [Pew Research Center: Google Users Are Less Likely to Click on Links When an AI Summary Appears in the Results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) - Evidence: Website study; Mar 2025 browsing · US adults - Sample: 900 adults; 68,879 unique searches reconstructed - Population: Google search visits in browsing histories of US adults in a tracking panel - Data period: Browsing March 1–31, 2025; search-result pages reconstructed April 7–17, 2025 - Source location: Clicking links within AI summaries - Limit: Observational comparison, with different query types in AI and non-AI groups. April reconstructions may differ from March results; US panel only. A browsing session ends when the user exits the browser for at least five seconds. - Evidence ID: S069; checked 2026-09-09 ### 98. People often stopped browsing after an AI summary **26% vs 16% of visits** — In Pew’s March 2025 US browsing study, 26% of Google visits with an AI summary ended the browsing session, versus 16% without one. **In plain English:** People were more likely to stop browsing when Google showed an AI summary. Different kinds of searches may explain part of the gap. **What you can do:** 1. Track which important Google searches show an AI summary and whether visits from those searches change. 2. Treat the pattern as a clue to investigate because the study compared different kinds of searches. [Pew Research Center: Google Users Are Less Likely to Click on Links When an AI Summary Appears in the Results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) - Evidence: Website study; Mar 2025 browsing · US adults - Sample: 900 adults; 68,879 unique searches reconstructed - Population: Google search visits in browsing histories of US adults in a tracking panel - Data period: Browsing March 1–31, 2025; search-result pages reconstructed April 7–17, 2025 - Source location: Browsing sessions ending after Google search - Limit: Observational comparison, with different query types in AI and non-AI groups. April reconstructions may differ from March results; US panel only. A browsing session ends when the user exits the browser for at least five seconds. - Evidence ID: S070; checked 2026-09-09 ### 99. Top-position CTR was lower than the model expected **58% lower estimated CTR** — Ahrefs’ February 2026 update estimated that AI Overviews were associated with 58% lower desktop position-one CTR than expected without them. **In plain English:** Top-ranking pages received fewer clicks than expected when AI Overviews appeared. This was an estimate, not a controlled test. **What you can do:** 1. Track AI Overviews and click rates for important desktop searches. 2. Use the pattern for forecasts, but do not promise the same traffic loss. [Ahrefs: AI Overviews Reduce Clicks by 58% (Updated Study)](https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/) - Evidence: Website study; 2026 release · modeled CTR comparison - Sample: 300,000 keywords, split into two groups of 150,000 - Population: Desktop Google Search Console data for AI Overview and non-AI informational keyword groups - Data period: December 2023 versus December 2025 - Source location: Updated CTR comparison and counterfactual method - Limit: Trend-adjusted counterfactual from observational groups, not a randomized test. The 58% estimate is relative to expected CTR without AI Overviews, not a raw before/after difference. - Evidence ID: S071; checked 2026-09-09 ### 100. Google still sent far more referral traffic **About 190× ChatGPT** — In Ahrefs’ February 2026 analysis of 76,000 websites, Google sent about 190 times as much referral traffic as ChatGPT. **In plain English:** In this website sample, Google sent far more visits than ChatGPT. AI referrals were still small beside search traffic. **What you can do:** 1. Show Google visits and ChatGPT visits on separate lines in your monthly report. 2. Keep investing in useful search pages while you test whether AI referrals are growing for your audience. [Ahrefs: ChatGPT Has 12% of Google’s Search Volume, But Google Sends 190x More Traffic to Websites](https://ahrefs.com/blog/chatgpt-has-12-percent-of-googles-search-volume/) - Evidence: Website study; Feb 2026 · 76,000 websites - Sample: 76,000 websites - Population: Websites using Ahrefs Web Analytics - Data period: not reported; February 2026 analysis - Source location: Google sends more traffic to websites; analytics comparison - Limit: Vendor analytics sample and undisclosed lookback period. Referral traffic is not search volume or market share. Rounded ratio is the publisher’s reported comparison. - Evidence ID: S072; checked 2026-09-09 ### 101. Web mentions correlated with AI visibility **0.66–0.71 correlation** — Ahrefs’ December 2025 analysis of 75,000 brands reported correlations of 0.66–0.71 between web mentions and AI brand visibility. **In plain English:** Brands mentioned more often online also appeared more often in AI answers. That does not prove that paid mentions will help. **What you can do:** 1. List accurate brand mentions on relevant news, industry and community sites. 2. Check which sources AI answers use and improve missing facts. [Ahrefs: Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied)](https://ahrefs.com/blog/ai-brand-visibility-correlations/) - Evidence: Website study; Dec 2025 release · 75,000 brands - Sample: 75,000 brands and millions of AI responses - Population: Brands identified from domains with DR above 40 and a top keyword with at least 800 monthly searches; AI responses in Brand Radar - Data period: Snapshot dates not reported; includes AI Overview results from earlier 2025 research - Source location: Key takeaways; brand mentions; methodology - Limit: Spearman correlations in an established-brand-biased sample; not ranking factors or evidence of causation. Web-mention and YouTube-mention measures partly overlap. Platform measurements are not all contemporaneous. - Evidence ID: S073; checked 2026-09-09 ### 102. YouTube mentions had a strong association **About 0.737 correlation** — In its December 2025 brand analysis, Ahrefs reported an approximately 0.737 correlation between YouTube mentions and AI visibility across the platforms studied. **In plain English:** YouTube mentions had a strong link with AI visibility. Well-known brands may simply be visible in both places. **What you can do:** 1. Make useful, well-sourced videos for questions that benefit from a demonstration. 2. Measure video mentions, AI citations and brand visibility separately. [Ahrefs: Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied)](https://ahrefs.com/blog/ai-brand-visibility-correlations/) - Evidence: Website study; Dec 2025 release · 75,000 brands - Sample: 75,000 brands and millions of AI responses - Population: Brands identified from domains with DR above 40 and a top keyword with at least 800 monthly searches; AI responses in Brand Radar - Data period: Snapshot dates not reported; includes AI Overview results from earlier 2025 research - Source location: Key takeaways; YouTube mentions show the strongest correlation - Limit: Spearman correlations in an established-brand-biased sample; not ranking factors or evidence of causation. Web-mention and YouTube-mention measures partly overlap. Platform measurements are not all contemporaneous. - Evidence ID: S074; checked 2026-09-09 ### 103. AI citations extended beyond the organic top ten **37.1% also in the top 10** — In Ahrefs’ March 2026 study, 37.1% of AI Overview citation URLs also ranked among the top ten organic blue links for the same query. **In plain English:** Most AI citation URLs were outside the normal top ten. AI tools can find sources beyond the first search page. **What you can do:** 1. Ask the same important questions regularly and record every cited page. 2. Study why sources outside the top search results still earn citations. [Ahrefs: Update: 38% of AI Overview Citations Pull From the Top 10](https://ahrefs.com/blog/ai-overview-citations-top-10/) - Evidence: Website study; Mar 2026 · 4 million citation URLs - Sample: 863,000 keyword SERPs and 4 million AI Overview URLs - Population: Same-query Google SERPs and AI Overview citation URLs in Ahrefs Brand Radar - Data period: not reported; March 2026 release - Source location: Organic position table, second test excluding ads and SERP features - Limit: Organic blue-link top ten and first ten mixed SERP blocks are different definitions. Citation parsing changed since July 2025, so older 76.1% cannot support a like-for-like time trend. YouTube non-ranking figure uses the publisher’s broader results framing. - Evidence ID: S075; checked 2026-09-09 ### 104. YouTube supplied some citations outside the top 100 **18.2% of that citation subset** — In Ahrefs’ March 2026 study, YouTube supplied 18.2% of AI Overview citation URLs absent from the top 100 results for the same keyword. **In plain English:** Some cited YouTube videos were not in the top 100 search results. Video can surface differently in AI answers. **What you can do:** 1. Review cited YouTube videos that rank poorly in normal search. 2. Note which speakers, topics and formats make them useful to AI answers. [Ahrefs: Update: 38% of AI Overview Citations Pull From the Top 10](https://ahrefs.com/blog/ai-overview-citations-top-10/) - Evidence: Website study; Mar 2026 · 4 million citation URLs - Sample: 863,000 keyword SERPs and 4 million AI Overview URLs - Population: Same-query Google SERPs and AI Overview citation URLs in Ahrefs Brand Radar - Data period: not reported; March 2026 release - Source location: 18% of non-ranking citations come from YouTube - Limit: Organic blue-link top ten and first ten mixed SERP blocks are different definitions. Citation parsing changed since July 2025, so older 76.1% cannot support a like-for-like time trend. YouTube non-ranking figure uses the publisher’s broader results framing. - Evidence ID: S076; checked 2026-09-09 ### 105. Paid content supplied a small citation share **0.3% of cited links** — Muck Rack’s May 2026 study classified 0.3% of cited links as paid or advertorial content. **In plain English:** Paid or sponsored pages made up a very small share of citations. Buying a placement does not guarantee AI visibility. **What you can do:** 1. Report paid placements and AI citations as separate results. 2. Check the same questions before and after publishing, without promising a change. [Muck Rack / Generative Pulse: What Is AI Reading? New insights for May 2026](https://muckrack.com/blog/what-is-ai-reading-new-insights-may) - Evidence: AI answer study; May 2026 release · constructed prompts - Sample: More than 25 million citation links, per the companion May methodology article - Population: AI responses to constructed, unbranded discovery prompts across 17 industries - Data period: May 2026 edition; collection window not disclosed in public write-ups - Source location: Paid media section; companion methods - Limit: Designed prompts are not a random sample of user queries. Transactional branded prompts were intentionally excluded. Model versions, answer settings and citation classification affect the result. The public summary repeats an older 1-million sample description; the companion methods explicitly gives 25 million for this edition. - Evidence ID: S136; checked 2026-09-09 ### 106. Citation frequency differed across AI systems **96% / 82% / 55%** — In Muck Rack’s May 2026 tested responses, 96% from ChatGPT, 82% from Gemini and 55% from Claude included citations. **In plain English:** The three AI systems cited sources at different rates. Results from one platform should not become a rule for another. **What you can do:** 1. Ask the same approved questions in ChatGPT, Gemini and Claude. 2. Record the date and compare each system with its own past results. [Muck Rack / Generative Pulse: What Is AI Reading? New insights for May 2026](https://muckrack.com/blog/what-is-ai-reading-new-insights-may) - Evidence: AI answer study; May 2026 release · constructed prompts - Sample: More than 25 million citation links, per the companion May methodology article - Population: AI responses to constructed, unbranded discovery prompts across 17 industries - Data period: May 2026 edition; collection window not disclosed in public write-ups - Source location: Citation behavior varies across platforms - Limit: Designed prompts are not a random sample of user queries. Transactional branded prompts were intentionally excluded. Model versions, answer settings and citation classification affect the result. The public summary repeats an older 1-million sample description; the companion methods explicitly gives 25 million for this edition. - Evidence ID: S137; checked 2026-09-09 ### 107. The question changes the source mix **46% of trend responses** — In Muck Rack’s May 2026 study, 46% of industry-trend responses cited journalism. **In plain English:** Trend questions cited journalism often. Different questions lead AI systems to different types of sources. **What you can do:** 1. Group questions by type, such as trends, comparisons or instructions. 2. Create the kind of source each question needs and check results again later. [Muck Rack / Generative Pulse: What Is AI Reading? New insights for May 2026](https://muckrack.com/blog/what-is-ai-reading-new-insights-may) - Evidence: AI answer study; May 2026 release · constructed prompts - Sample: More than 25 million citation links, per the companion May methodology article - Population: AI responses to constructed, unbranded discovery prompts across 17 industries - Data period: May 2026 edition; collection window not disclosed in public write-ups - Source location: Question type and journalism citations - Limit: Designed prompts are not a random sample of user queries. Transactional branded prompts were intentionally excluded. Model versions, answer settings and citation classification affect the result. The public summary repeats an older 1-million sample description; the companion methods explicitly gives 25 million for this edition. - Evidence ID: S138; checked 2026-09-09 ## Original-source index - [Ahrefs — Google Says “Links Matter Less”—We Looked at 1,000,000 SERPs to See if It’s True](https://ahrefs.com/blog/links-matter-less-but-still-matter/) · not reported; published January 2025 - [Ahrefs — 96.55% of Content Gets No Traffic From Google. Here’s How to Be in the Other 3.45%](https://ahrefs.com/blog/search-traffic-study/) · not reported; 2023 report - [Ahrefs — Reciprocal Links: Do They Help or Hurt Your SEO?](https://ahrefs.com/blog/reciprocal-links/) · not reported - [BuzzStream — Cost of Guest Posts, Based on 500k Sites [Updated 2026 Data]](https://www.buzzstream.com/blog/guest-post-costs/) · June 2026 release; collection dates undisclosed - [Aira — 11 Takeaways from the 2022 State of Link Building Report](https://aira.net/blog/11-takeaways-2022-state-of-link-building-report/) · 16 March–2 May 2022 - [Reboot Online — Digital PR Statistics](https://www.rebootonline.com/digital-pr-statistics/) · Agency coverage acquired during 2024 - [BuzzStream and Citation Labs — Link Building Trends Report (2025)](https://www.buzzstream.com/blog/link-building-trends/) · not reported; 2025 release - [BuzzStream — State of Digital PR Report (2026)](https://www.buzzstream.com/blog/state-of-digital-pr-2026/) · not reported; 2026 release - [BuzzStream — How Much Does Digital PR Cost?](https://www.buzzstream.com/blog/digital-pr-costs/) · 2025 survey; fieldwork dates not disclosed; page updated June 2026 - [Backlinko and Pitchbox — We Analyzed 12 Million Outreach Emails. Here’s What We Learned](https://backlinko.com/email-outreach-study) · not reported; 2019 study - [Reboot Online — Digital PR Statistics: email subject-line analysis](https://www.rebootonline.com/digital-pr-statistics/) · Collection period not reported - [Muck Rack — The State of Journalism 2026](https://media.muckrack.com/documents/State_of_Journalism_2026_1.pdf) · 2026-01-30 to 2026-03-02 - [Cision — Why Do Journalists Ignore Your PR Pitch?](https://www.cision.com/resources/articles/why-do-journalists-ignore-your-pr-pitch/) · not reported; 2026 survey release - [Orbit Media — Annual Blogger Survey 2025](https://www.orbitmedia.com/blog/blogging-statistics/) · August 2025 - [Ahrefs — Marketers Using AI Publish 42% More Content](https://ahrefs.com/blog/marketers-using-ai-publish-more-content/) · December 2024–February 2025 - [Backlinko / BuzzSumo — We Analyzed 912 Million Blog Posts. Here’s What We Learned About Content Marketing](https://backlinko.com/content-study) · Articles published 1 October 2017–1 October 2018 - [Ahrefs — Link Building Case Study: How We Built Backlinks With a Stats Page](https://ahrefs.com/blog/link-building-case-study/) · Campaign reported August 2020; precise outreach window undisclosed - [Siege Media — Organic Link Ratios by Industry: What 100 Sites Reveal About Link Acquisition](https://www.siegemedia.com/research/organic-link-ratios) · January 2024–December 2025 - [Ahrefs — At Least 66.5% of Links to Sites in the Last 9 Years Are Dead](https://ahrefs.com/blog/link-rot-study/) · Links tracked from January 2013 across a described nine-year window; exact endpoint not reported - [Pew Research Center — When Online Content Disappears](https://www.pewresearch.org/data-labs/2024/05/17/when-online-content-disappears/) · 2013–2023 Common Crawl pages checked October 2023; news, government and Wikipedia samples from spring 2023 - [Ahrefs — We Crawled Over 1 Million Domains to Find the Most Common Technical SEO Issues](https://ahrefs.com/blog/site-audit-study/) · not reported; August 2023 report - [Zyppy — 23 Million Internal Links – SEO Case Study](https://zyppy.com/seo/seo-study/) · not reported - [SearchPilot — SEO Split Test Lessons: Increasing Internal Linking](https://www.searchpilot.com/resources/case-studies/seo-split-test-lessons-increasing-internal-linking) · not reported; case study published August 2020 - [SearchPilot — SEO Split Test Lessons: Reducing a Link Block](https://www.searchpilot.com/resources/case-studies/seo-split-test-lessons-reducing-link-block) · not reported; case study published April 2024 - [WebAIM — The WebAIM Million: The 2026 Report on the Accessibility of the Top 1,000,000 Home Pages](https://webaim.org/projects/million/) · February 2026 - [SearchPilot — Does adding internal links to your home page’s footer improve organic traffic?](https://www.searchpilot.com/resources/case-studies/adding-internal-links-to-home-page-footer) · Test reported September 2021; exact dates not reported - [SearchPilot — Will removing self-referential breadcrumb links improve organic traffic?](https://www.searchpilot.com/resources/case-studies/removing-self-referential-breadcrumb) · Test reported July 2024; exact dates not reported - [Pew Research Center — Google Users Are Less Likely to Click on Links When an AI Summary Appears in the Results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) · Browsing March 1–31, 2025; search-result pages reconstructed April 7–17, 2025 - [Ahrefs — AI Overviews Reduce Clicks by 58% (Updated Study)](https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/) · December 2023 versus December 2025 - [Ahrefs — ChatGPT Has 12% of Google’s Search Volume, But Google Sends 190x More Traffic to Websites](https://ahrefs.com/blog/chatgpt-has-12-percent-of-googles-search-volume/) · not reported; February 2026 analysis - [Ahrefs — Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied)](https://ahrefs.com/blog/ai-brand-visibility-correlations/) · Snapshot dates not reported; includes AI Overview results from earlier 2025 research - [Ahrefs — Update: 38% of AI Overview Citations Pull From the Top 10](https://ahrefs.com/blog/ai-overview-citations-top-10/) · not reported; March 2026 release - [Muck Rack / Generative Pulse — What Is AI Reading? New insights for May 2026](https://muckrack.com/blog/what-is-ai-reading-new-insights-may) · May 2026 edition; collection window not disclosed in public write-ups ## Refresh plan Recheck links and fast-changing AI/marketplace evidence by 9 December 2026. Review annual surveys when new editions appear; keep prior-year observations visibly dated. Preserve stable evidence IDs, re-read changed methods, and do not stitch incompatible editions into a trend. Hold a claim if its source becomes unavailable, its numerator or denominator is ambiguous, or conflicting values cannot be resolved. Recalculate the visible count after any removal. See refresh-plan.md and the full editorial evidence log for the maintenance workflow.