AI search statistics,
with their receipts.
Every statistic Citevera cites anywhere on this site, in our reports, and in our plugin - with its source, sample size, date, later revisions, and an honest note on how much weight it can bear. If a number here is correlational, we say so. If it comes from a vendor, we say that too.
Written by Paul, founder of Citevera · Sources verified 2026-08-04 · Last reviewed 2026-08-04
How we frame these numbers
- Independent measurement - a non-vendor organization observing real user behavior. The strongest evidence tier; we lead with it.
- Vendor analysis - a tool vendor analyzing its own large dataset. Valuable and usually the only data at this scale, but the vendor sells a product adjacent to the finding.
- Vendor data, correlational - an association measured on one platform. Sites that differ on the measured signal differ in many other ways too; these numbers never support causal claims here.
- Projection - a forward-looking estimate, labelled as one.
When an AI summary appears, users click a search result 8% of the time vs 15% without one; only ~1% click sources inside the AI answer.
Our framing: Measured behavior from a real-user panel, not a vendor dataset - the most independent evidence available on AI-search click behavior.
Pew Research Center, July 2025, panel of ~900 US adults' actual browsing · Primary source
38% of AI Overview citations come from Google's top-10 organic results. Ahrefs measured 76% in 2025; the 2026 follow-up found 38% - engines now cite well beyond the top 10, though ranking position remains the strongest single predictor of being cited.
Our framing: Vendor analysis of AI Overview citation sources. Use for citation mechanics (authority still matters, but is no longer the whole story), not for traffic claims.
Ahrefs baseline, 2025: Ahrefs' original 2025 analysis (1.9M citations across ~300k keywords) measured 76% - the baseline the 2026 figure halved. Original study
Ahrefs, 2026, 863k SERPs · Primary source
The top organic result saw ~34.5% fewer clicks when an AI Overview appeared.
Our framing: Vendor CTR analysis. Note Ahrefs and Semrush disagree on the overall zero-click direction; on that specific sub-point cite both or neither.
Ahrefs follow-up, May 2026: Ahrefs' May 2026 update measured a 58% reduction - the effect has grown, not faded. Follow-up study
Ahrefs, March 2025, 300k keywords (150k with AI Overviews, 150k without) · Primary source
AI-search visitors converted at ~4.4x the rate of traditional organic visitors - fewer clicks, but the ones that arrive are further down the funnel.
Our framing: Vendor analysis. Semrush's dataset conflicts with Ahrefs' on the zero-click direction - do not use either alone for that sub-point. Semrush's forward estimates (AI channels matching organic value by ~2027-2028) are projections, not measurements.
Semrush, 2025, 10M+ keyword dataset · Primary source
Sites that allowed AI crawling saw 3.2x more traffic (mean 527.7 vs 164.9 sessions per site; the published study does not state the measurement window) and 2.7x more form submissions (4.17 vs 1.57) than sites that did not.
Our framing: Single-platform, correlational: sites that allow AI crawlers differ in many ways from sites that block them. Use for local-SMB context only; never as the sole citation for a causal claim.
- 81% / 16.6% / 1.8% / 0.6%GPTBot accounted for 81% of AI crawler traffic, ClaudeBot 16.6%, PerplexityBot 1.8%, Gemini 0.6%.
- 33xSites with 50+ blog posts averaged 1,373.7 AI crawler visits vs 41.6 for sites with no blog - a 33x gap.
- 92.8% vs 58.9%Sites with Google Business Profile sync saw a 92.8% AI crawl rate vs 58.9% without.
- 89.8%Sites with third-party review-platform integrations saw an 89.8% AI crawl rate and averaged 376.9 crawler visits.
Duda, February 2026, 858k sites on a single platform · Primary source
So - is AEO real?
Answer engine optimization is real in the ways that are measured, and oversold in the ways that are not.
What is established: user behavior has shifted. When an AI summary appears, users click a search result 8% of the time vs 15% without one; only ~1% click sources inside the AI answer. That is an independent panel measuring real browsing, not a vendor deck. And the mechanics of who gets cited are not random: 38% of AI Overview citations come from Google's top-10 organic results. Ahrefs measured 76% in 2025; the 2026 follow-up found 38% - engines now cite well beyond the top 10, though ranking position remains the strongest single predictor of being cited.
What is vendor-measured but plausible: AI-search visitors converted at ~4.4x the rate of traditional organic visitors - fewer clicks, but the ones that arrive are further down the funnel. Semrush and Ahrefs disagree on the overall zero-click direction, which is exactly why we cite the specific finding and not the narrative.
What is correlational: Sites that allowed AI crawling saw 3.2x more traffic (mean 527.7 vs 164.9 sessions per site; the published study does not state the measurement window) and 2.7x more form submissions (4.17 vs 1.57) than sites that did not. Those are single-platform correlations on local-SMB-heavy sites - useful for knowing which signals crawlers reward, not proof that flipping a switch multiplies your traffic.
The defensible conclusion: being findable, parseable, and citable is increasingly important as AI-mediated discovery grows, the click economics have genuinely changed, and most causal "AI visibility grew our traffic 3x" claims outrun their data. Optimize on the established parts; treat the rest as upside.