AEO.wikiThe AEO Reference

AEO research and studies

From AEO.wiki, the Answer Engine Optimization reference · Published · Updated · By the · How we research

The strongest evidence so far comes from two randomized field experiments run in early 2026: AI Overviews cut outbound organic clicks by 39.8%,[11] and an AI Mode-only experience cut click-through by 18.8 percentage points,[12] and neither study found a better user experience. Observational data points the same way.[16][15][9] AI Overview citations now overlap far less with the top 10 (38% in March 2026, down from 76% in July 2025),[13][7] AI assistants’ news answers had at least one significant issue 45% of the time,[19] and a peer-reviewed 2026 replication found that the GEO paper’s text edits (adding citations, quotes and statistics) did not beat unmodified pages.[20] Every study has limits, so each is listed below with its method, date and caveats.

Studies at a glance

Study (date)MethodKey finding and caveats
Randomized field experiments
Agarwal and Sen field experiment (SSRN, Apr 2026; revised Jul 2026)[11]Preregistered randomized trial using a Chrome extension: 1,065 U.S. users, 68,089 searches, Jan 7–Feb 10, 2026Conditional on appearing, AI Overviews cut outbound organic clicks by 39.8% and raised zero-click searches by 34.5%, without affecting sponsored clicks or search frequency; no measurable improvement in perceived search quality. Working paper, not peer reviewed.
Wang, Gleason, Bart, Wilson and Metaxa (arXiv, Aug 2026)[12]Preregistered field experiment on Google Search, N=1,100 U.S. participants recruited March 2026; 3 baseline and 7 treatment daysAn AI Mode-only experience cut click-through by 18.8 percentage points and lowered trust (−0.34 on a 7-point scale); removing AI Overviews raised click-through by 8.8 points. Preprint. Only 51.1% of AI Overviews were successfully hidden; sample is younger, more educated and more left-leaning than the U.S. population.
Clicks and traffic (observational)
Khosravi and Yoganarasimhan (arXiv v6, Sep 2026)[16]Difference-in-differences on Wikipedia clickstream data, Dec 2023–Dec 2024: English articles vs German and French equivalentsAI Overviews cut English Wikipedia search referrals by 5.45% (vs German) and 4.82% (vs French), an estimated 100.27 million fewer visits a month. Working paper; reduced-form estimate.
Ahrefs, France launch (Sep 2026)[15]Search Console data for 963 domains, 28 days before vs 9 days after AI Overviews launched in France on Jul 22, 2026Domains where more than 20% of queries showed an AI Overview saw median CTR fall 23.1% (24.8% after adjusting for a control group); 82.2% of them lost CTR. Short window; vendor study.
Seer Interactive (Apr 2026)[14]53 brands, 5.47M queries, 2.43B organic impressions, Jan 2025–Feb 2026Organic CTR on AI Overview queries rose from 1.3% (Dec 2025) to 2.4% (Feb 2026). On informational queries, brands cited in the AI Overview got 120% more organic clicks per impression than uncited brands, but 38% fewer than on queries without an AI Overview. Correlational; Seer says it cannot claim causation.
Ahrefs CTR update (Feb 2026)[9]December 2025 dataTop-result CTR 58% lower when an AI Overview appears. Vendor study.
Pew Research Center (Jul 2025)[6]Browsing data from 900 U.S. adults, March 2025; 68,879 searches18% of searches showed an AI summary; users clicked a result in 8% of visits with a summary vs 15% without; 1% clicked a link inside the summary. Dated data (March 2025); see [17] for 2026 survey data.
Ahrefs CTR (Apr 2025)[5]300,000 keywords, March 2025 dataAI Overviews associated with a 34.5% lower CTR for the top result. Historical: superseded by [9].
Citations and visibility
Ahrefs citations update (Mar 2026)[13]863,000 keyword SERPs; 4 million AI Overview URLs from Ahrefs Brand Radar37.9% of cited URLs ranked in the top 10, 31.2% in positions 11–100 and 31.0% beyond 100. Ahrefs links the drop from about 76% to query fan-out after AI Overviews moved to Gemini 3; its parsing method also changed. Vendor study.
Ahrefs citations (Jul 2025)[7]AI Overview citations compared with rankings76.1% of cited pages ranked in the top 10. Historical figure; superseded by [13].
Ahrefs brand visibility (Dec 2025)[8]75,000 brands; mentions in ChatGPT, AI Mode, AI OverviewsYouTube mentions (~0.74) and branded web mentions (~0.66–0.71) correlated most with AI visibility
SparkToro and Gumshoe (Jan 2026)[21]600 volunteers ran 12 prompts 2,961 times in ChatGPT, Claude and Google AI Overviews/AI Mode, Nov–Dec 2025Less than a 1-in-100 chance of getting the same list of brands twice, and about 1 in 1,000 of getting the same order; how often a brand appears is a more reasonable metric than its position. Not peer reviewed; Gumshoe sells AI tracking (disclosed).
Ahrefs llms.txt study (Jun 2026)[22]All 137,210 domains in Ahrefs Web Analytics with traffic in May 2026, plus bot logs28% of domains published llms.txt; 97% of those files received zero requests. Of the requests that did arrive, 96% came from bots and 1.1% from AI retrieval bots. Sample skews SEO-aware, so 28% is an upper bound; vendor study.
Vercel and MERJ: The rise of the AI crawler (Dec 2024)[3]Crawler traffic across Vercel’s networkGPTBot (569M fetches in a month) and Claude (370M) did not execute JavaScript; Google and Apple did. Dated: December 2024 data; crawler behavior may have changed.
Usage surveys
Pew Research Center (Jun 2026)[17]Survey of 5,119 U.S. adults, Feb 17–23, 2026About half of U.S. adults use AI chatbots, up from a third in 2024; 42% use them to search for information; 60% read AI summaries at the top of search results. ChatGPT (44%) is used far more than Gemini (24%) or Copilot (17%). Self-reported.
Reuters Institute Digital News Report (Jun 2026)[18]Online survey of 97,520 people in 48 markets, early 2026Weekly use of AI chatbots for news rose from 7% to 10%; 1% say AI is their main news source. Across 27 markets, 4% always or often click through to sources from AI, vs 19% from search. Self-reported.
Answer accuracy
EBU and BBC, News Integrity in AI Assistants (Oct 2025)[19]22 public service media organizations in 18 countries and 14 languages rated 2,709 answers from ChatGPT, Copilot, Gemini and Perplexity (May–Jun 2025)45% of answers had at least one significant issue; sourcing was the biggest cause (31%), then accuracy (20%). Gemini had significant issues in 76% of responses, Copilot 37%, ChatGPT 36%, Perplexity 30%.
Tow Center (Mar 2025)[4]1,600 queries across 8 AI search tools, asking them to identify news article sourcesMore than 60% of answers were incorrect; Perplexity 37% wrong, Grok 3 94%. Superseded by the larger EBU/BBC study [19].
Content-optimization benchmarks
FeatGEO (ACL 2026)[20]GEO-Bench with GPT-4o-mini, Gemini 2.5 Flash and Qwen-plus; re-tested the GEO paper’s nine text heuristics against a feature-level methodAll nine GEO heuristics, including Cite Sources, Quotation Addition and Statistics Addition, scored below the unmodified page on all three engines (for example 10.92–12.21% vs 13.34% visibility on GPT-4o-mini); feature-level rewriting reached 18.31%. Peer reviewed; a benchmark, not live search.
GEO: Generative Engine Optimization (Nov 2023; KDD 2024)[1]GEO-bench of diverse queries; tested content changes against generative enginesMethods such as adding citations, quotations and statistics raised source visibility by up to 40%; effects varied by domain. Gains not reproduced by FeatGEO [20].
Ranking manipulation for conversational search (EMNLP 2024)[2]Adversarial text injected into product pagesPrompt injection can reorder which products LLM search recommends, including on a production engine

What the evidence supports

How to read AEO studies

  1. Check the date: engines change monthly.
  2. Check who ran it. Vendor studies often promote the vendor’s tools.
  3. Correlation is not causation.
  4. Prompt-based studies sample outputs; they don’t see real query volumes.
  5. Check the design: only two studies here are randomized trials;[11][12] the rest are observational, surveys or benchmarks.
  6. Check review status: working papers and preprints can change between versions.

More debunked claims are on AEO myths. Every source used on this site is listed on sources.

Where studies conflict, and why

Several figures above look inconsistent until you compare dates, samples and what was measured.

What the studies do not measure

Our reading: none of the studies listed here is a controlled test of one tactic (such as schema markup, llms.txt or a rewrite) on live AI citations, so any claim that a tactic raises citations by a fixed percentage on your site is unproven. Measure your own pages instead.

Using this evidence on your own site

  1. Confirm eligibility first. Google requires a page to be indexed and snippet-eligible to appear as a supporting link in its AI features.[10]
  2. Use first-party data. Search Console’s generative AI performance report and Bing’s AI Performance report show real impressions and citations (see measurement).
  3. Track mention frequency, not rank. Brand lists in AI answers rarely repeat in the same order.[21]
  4. Treat sourced, verifiable writing as good practice, not a proven lever.[20]
  5. Re-check dates. Engines change monthly, and the table above lists each study’s date and caveats.

Frequently asked questions

Do AI Overviews reduce clicks?

Yes. In a randomized trial, AI Overviews cut outbound organic clicks by 39.8% (Agarwal and Sen, 2026), and a second trial found an AI Mode-only experience cut click-through by 18.8 percentage points (Wang et al., 2026). Neither is peer reviewed yet. Observational data agrees: Ahrefs estimated a 58% drop in top-result CTR (December 2025 data).

Is there proof AEO works?

Not yet. The 2024 GEO paper showed gains in a benchmark, but a peer-reviewed ACL 2026 replication (FeatGEO) found those text edits (citations, quotes, statistics) scored below unmodified pages. Real-world evidence is mostly correlational, so test on your own site.

Do AI Overviews still cite top-ranking pages?

Less than before. Ahrefs found 37.9% of cited URLs ranked in the top 10 in March 2026, down from 76.1% in July 2025; 31.0% ranked beyond position 100.

References

Pages accessed September 28, 2026 (references 11–22: September 30, 2026) unless a date is given. See all sources and our editorial policy.

  1. ^ "GEO: Generative Engine Optimization (arXiv:2311.09735; Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan, Deshpande)". arXiv / KDD 2024. Submitted November 16, 2023; revised June 28, 2024. Gains not reproduced by [20].
  2. ^ "Ranking Manipulation for Conversational Search Engines (Pfrommer, Bai, Gautam, Sojoudi)". Proceedings of EMNLP 2024, ACL Anthology. Published November 2024.
  3. ^ "The rise of the AI crawler". Vercel (with MERJ). Published December 17, 2024. Dated: crawler behavior may have changed since.
  4. ^ "AI Search Has a Citation Problem (Jaźwińska and Chandrasekar)". Columbia Journalism Review, Tow Center for Digital Journalism. Published March 6, 2025. Superseded by the larger EBU/BBC study ([19]).
  5. ^ "AI Overviews Reduce Clicks by 34.5%". Ahrefs. Published April 17, 2025; March 2025 data. Historical: superseded by [9].
  6. ^ "Google users are less likely to click on links when an AI summary appears in the results". Pew Research Center. Published July 22, 2025; March 2025 browsing data. Dated data; see [17] for 2026 survey data.
  7. ^ "76% of AI Overview Citations Pull From the Top 10". Ahrefs. Published July 21, 2025. Historical figure; superseded by [13].
  8. ^ "AI brand visibility correlations (75,000 brands)". Ahrefs. Published December 12, 2025.
  9. ^ "Update: AI Overviews Reduce Clicks by 58%". Ahrefs. Published February 4, 2026; December 2025 data.
  10. ^ "Optimizing your website for generative AI features on Google Search". Google Search Central. Added May 15, 2026; last updated July 10, 2026.
  11. ^ "The Impact of Google AI Overviews on Publisher Traffic and User Experience: Evidence from a Field Experiment (Agarwal and Sen)". SSRN working paper (Indian School of Business; Carnegie Mellon University). Posted April 3, 2026; revised July 8, 2026. Not peer reviewed.
  12. ^ "AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence (Wang, Gleason, Bart, Wilson, Metaxa)". arXiv preprint (University of Pennsylvania; Northeastern University). Submitted August 18, 2026. Not peer reviewed.
  13. ^ "Update: 38% of AI Overview Citations Pull From The Top 10 (Louise Linehan)". Ahrefs. Published March 2, 2026. Vendor study.
  14. ^ "AIO Impact on Google CTR: 2026 Update (McDonald, Cooley, Williams)". Seer Interactive. Published April 24, 2026; January 2025–February 2026 data. Correlational.
  15. ^ "AI Overviews Cut CTR by 23.1% in France (Juliette Begue)". Ahrefs. Published September 16, 2026. Vendor study; 9-day post-launch window.
  16. ^ "Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia (Khosravi and Yoganarasimhan)". arXiv working paper (University of Washington). Version 6, September 2, 2026. Not peer reviewed.
  17. ^ "Americans and AI 2026: Chatbots, Smart Devices and Views on Impact". Pew Research Center. Published June 17, 2026; survey of 5,119 U.S. adults, February 17–23, 2026.
  18. ^ "Emerging uses of AI chatbots for news and what it means for journalism (Amy Ross Arguedas), Digital News Report 2026". Reuters Institute for the Study of Journalism. Published June 16, 2026; online survey, early 2026. Self-reported.
  19. ^ "News Integrity in AI Assistants". European Broadcasting Union (EBU) and BBC. Published October 22, 2025; responses generated May–June 2025.
  20. ^ "Think Before Writing: Feature-Level Multi-Objective Optimization for Generative Citation Visibility (Liu and Xu)". Proceedings of ACL 2026 (Long Papers), ACL Anthology. Pages 20290–20303. Peer reviewed.
  21. ^ "NEW Research: AIs are highly inconsistent when recommending brands or products; marketers should take care when tracking AI visibility (Rand Fishkin, with Patrick O’Donnell of Gumshoe)". SparkToro. Published January 28, 2026. Not peer reviewed; Gumshoe sells AI-visibility tracking.
  22. ^ "We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read (Louise Linehan)". Ahrefs. Published June 15, 2026; May 2026 data. Vendor study.
Categories: AEO research