Why We Ran This
Every proposal in this industry leans on statistics. The problem is that almost nobody traces them. A number appears in one article, gets quoted in a second. Also, by the tenth citation the original sample size, measurement window. Also, vertical have all fallen away. What remains is a bare percentage that sounds authoritative and means very little.
We started this audit for an unflattering reason. During a review of our own site we found a statistic on one of our pages attributed to the wrong source. The numbers were real. The attribution was not. We fixed it, then went looking for how widespread the problem was. It is widespread.
What follows is every figure we could trace, what it actually says. Also, where it does not say what people think it says. Where a claim failed, we say so. Where our own copy was wrong, we say that too.
The Findings, Summarized
| The claim as usually quoted | Verdict | What is actually true |
|---|---|---|
| 76% of AI Overview citations come from top 10 pages | Stale | That was Ahrefs' July 2025 figure. Their updated study puts it at 38%. BrightEdge measures lower still. |
| AI traffic converts at 14.2% vs 2.8%, per Adobe | Misattributed | Real figures, wrong source. They belong to the Opollo 2026 benchmark of 312 B2B firms, reproduced by RankScience. |
| Organic CTR fell 61% because of AI Overviews | Incomplete | True for June 2024 to September 2025. Seer's own 2026 update then recorded a rebound from 1.3% to 2.4%. |
| Adobe says AI traffic converts better | Holds | 42% better in March 2026, 54% by May. Retail only, vendor stated, measured against all non-AI traffic. |
| AI referrals drove 23x more signups | Holds, narrow | Ahrefs' own data, June 2025. A single B2B software company, not a cross-industry benchmark. |
| Schema markup lifts AI citations 4.8x | Untraceable | No primary publication located. This one was on our own site. It has been removed. |
| FAQ schema increases citations 2.3x | Untraceable | No primary publication located for the cited source. Also removed from our site. |
The Detail
This is the most consequential stale number in the industry, because it underpins the argument that AI visibility follows automatically from good rankings.
Ahrefs published that finding in July 2025. Their updated analysis, covering 863,000 keywords and 4 million AI Overview URLs, puts the figure at 38 percent, with the remainder split almost evenly between pages ranking 11 to 100 and pages that do not rank in the top 100 at all. BrightEdge, using different method, measures the overlap around 17 percent.
Ahrefs is candid that some of the change reflects improved citation detection in their own tooling rather than pure behavior change. So the two studies are not perfectly comparable. That caveat deserves to travel with the number. The direction, but, is not in dispute across independent datasets. Also, the practical implication is major: a meaningful share of AI citations now goes to pages with little or no traditional organic visibility. AI visibility has become something you can win separately. Also, something you can lose while your rankings look healthy.
The figures are real. The attribution is wrong, and it is wrong in a lot of places.
They come from the Opollo 2026 AI Search Benchmark Report, which analyzed GA4 referral data and CRM attribution across 312 B2B technology firms in North America, Australia, and the UK. RankScience reached the same numbers independently in a separate analysis, which is the strongest thing that can be said for any statistic.
Adobe published something different
Adobe published something different. Adobe Analytics measured AI referred traffic converting 42 percent better than non AI traffic in March 2026, rising to 54 percent by May, drawn from over a trillion visits across 130 plus North American retailers. Two caveats belong with it: the comparison is against all non AI traffic rather than Google organic specifically. Also, it is vendor stated rather than independently audited. Shopify's separately sourced commerce data points the same direction, reporting AI referred sessions converting close to 50 percent higher on product detail pages.
None of this is a small distinction. A B2B software benchmark and a retail benchmark answer different questions. Also, a business deciding a budget deserves to know which one it is looking at.
This is the most quoted number in search marketing right now. Also, the quoting has frozen it in time.
Seer Interactive measured organic CTR on AI Overview queries falling from 1.76 percent to 0.61 percent between June 2024 and September 2025, across 3,119 informational queries and 42 organizations. It is rigorous work and the finding was real.
Seer's own 2026 update, covering 53 brands, 5.47 million queries, and 2.43 billion organic impressions, then recorded a rebound: organic CTR on AI Overview queries climbing from a floor of 1.3 percent in December 2025 to 2.4 percent in February 2026. Seer characterizes this as leveling off rather than recovery. Either way, anyone still quoting the 61 percent decline without the rebound is describing a floor that has since lifted.
Two further pieces context rarely travel
Two further pieces of context rarely travel with the headline. Estimates of the decline range from roughly 15 percent in the most conservative study to over 60 percent in the most aggressive, depending entirely on method and query set. And CTR fell on queries without AI Overviews too, which means the shift is broader than any single search feature.
These were on our own site. We could not locate a primary publication for either, from the cited sources or anywhere else. Both have been removed rather than restated.
The underlying advice was never wrong. Structured data does make a page's facts machine readable. Also, a question with a direct answer beneath it is genuinely the easiest structure for a model to quote. Those statements are defensible on their mechanics. The multipliers attached to them were not. Also, attaching a false precision to true advice is its own kind of dishonesty.
Five More Figures Worth Knowing, With Their Caveats Attached
These held up to tracing. Each is reported here with the context that usually gets stripped away.
ChatGPT carries the overwhelming majority of AI referral volume
Conductor's tracking puts roughly 87 percent of measurable AI referral traffic flowing through ChatGPT. The practical consequence is that treating AI search as one undifferentiated channel misallocates effort. Optimizing purely for Google AI Overviews addresses a minority of the referral volume actually arriving.
Roughly 28 percent of ChatGPT's most cited pages have no Google organic visibility
Ahrefs measured this in late 2025, and it is the finding with the largest strategic implication on this page. It means a page can be cited heavily by a major AI system while ranking nowhere in Google. Any visibility measurement built only on rank tracking is structurally blind to that traffic.
Google's own two AI surfaces barely overlap
Ahrefs found about 13.7 percent URL overlap between AI Overviews and AI Mode in December 2025, despite both being Google products. SE Ranking's earlier analysis found comparably low overlap. Being cited in one is not evidence of being cited in the other. Also, reporting that conflates them is measuring less than it claims.
Most marketers are not measuring this channel at all
Industry surveys put the share of marketers tracking AI search as a separate channel at roughly 14 percent. Most AI referral traffic lands in analytics as direct or generic referral, which means a great deal of reported channel performance is quietly wrong in both directions. Google Analytics added a native AI Assistant channel grouping in May 2026. This will improve this going forward but does nothing for historical data.
Publishers took the hardest measurable hit
Press Gazette and Chartbeat data show global Google search referral traffic to publishers fell by roughly a third in the year to November 2025, with United States publishers hit harder. This is the clearest evidence that the change is structural rather than cyclical. It is also publisher data. Also, a local service business does not have a publisher's traffic profile. So it should inform your thinking without setting your hopes.
How to Check a Statistic Yourself in Two Minutes
You do not need our audit to do this. The method is simple enough to run on any number in any proposal you receive.
First, find the primary source, not an article citing one. If three searches only surface blogs quoting each other, that is your answer. Second, find the sample: how many sites, how many queries, over what period. A percentage without a denominator is a rumor with a decimal point. Third, find the comparison group, because most misleading statistics are technically accurate against a comparison nobody stated. Fourth, check the date, and check whether the publisher has since updated it. That last step is the one almost nobody takes. Also, it is the step that caught the largest error in this audit.
If the number survives all four, use it. If it does not, ask the agency quoting it where they got it. The quality of that answer tells you more about the engagement ahead than any case study will.
What We Changed on Our Own Site
Publishing this without disclosing our own corrections would defeat the point. During the review that prompted this audit we found and fixed the following on portlandpeakseo.com: a conversion statistic attributed to Adobe that belongs to Opollo, a citation figure quoted at 76 percent that its publisher has since updated to 38 percent, two multiplier claims for which no primary source could be located. Also, self referential review markup that does not meet Google's structured data guidance for review rich results.
We also found eight pages whose structured data contained a trailing comma, which made the JSON invalid and meant those pages were publishing no machine readable data at all despite appearing to. That last one is a useful reminder that a schema block can look complete and do nothing. Also, it is worth validating your own.
Methodology
For each claim we attempted to reach a primary publication that names its sample size, measurement window, and method. Where a primary source existed, we recorded what it actually measured, including the vertical and the comparison group, because those two details are what most secondary citations discard.
Where a figure only appeared in secondary articles quoting one another, and no primary publication could be located, we marked it untraceable. Untraceable does not mean fabricated. It means it does not meet the bar for use in a document that asks somebody to spend money.
We make no claim to have audited every statistic in circulation. This covers the ones we meet most often, including the ones we had used ourselves.
How to Use Any of This
Directionally, and never as a forecast. Every figure here comes from a specific sample in a specific window in a specific vertical. Retail conversion data does not predict a law firm's results. B2B software benchmarks do not predict a plumber's. The numbers describe which way the ground is moving, not where your business will land.
The only figure that describes you is your own baseline, recorded before work starts. This is the argument for measuring your own position rather than borrowing someone else's. That is what our free SEO audit exists to do, and the findings are yours whether or not you ever hire us.
If you are evaluating proposals right now, three questions do most of the filtering. Ask where each statistic comes from and what it measured. Ask what your baseline is and who records it. Also, ask what happens in a month where nothing moves. Vagueness on any of the three is the answer.
Corrections and Contact
If any finding here is wrong, we want to know. Also, we will publish the correction with the same prominence as the original. That is the standard we are asking of everyone else, so it applies here first. Reach us at PortlandPeakSEO@gmail.com or (971) 280-2861.
This research is free to cite and quote with attribution and a link. Journalists and analysts are welcome to contact us for the underlying trace notes.
More background: what AI SEO is, how it differs from traditional SEO, how our process works, our AI SEO services, published pricing, and live client sites you can verify yourself.