How much of search is actually AI now? The real numbers
Semrush's own 10-million-keyword tracker put AI Overview prevalence at roughly 15.7% in November 2025. Other trackers claim 48-60% for the same month. Both can't be right, and the gap tells you more than either number does.
Published September 21, 2026
Ask five SEO blogs what percentage of Google searches now trigger an AI Overview and you will get five different numbers, usually somewhere between 15% and 60%. That is not a rounding difference. It is a four-times spread on the same underlying question, in the same rough time window. One of two things is true: either most of these sources are unreliable, or they are answering slightly different questions and presenting them as the same number. It's mostly the second, and the gap itself is worth understanding before you act on any single figure.
The number that's actually measured, not estimated
Semrush tracked the same panel of over 10 million keywords through 2025, the closest thing available to a consistent, methodology-disclosed baseline. AI Overview prevalence in that tracked set climbed from about 6.5% in January 2025 to a peak near 24.6% in July, then settled back to roughly 15.7% by November as Google pulled overviews back on some query types.
Of the searches that did trigger an AI Overview in October 2025, 57.1% carried informational intent, against 42.9% commercial, transactional, or navigational. That split matters more than the headline percentage, because it explains almost every other number you'll see quoted.
Where the 43-60% figures come from
The higher numbers circulating in SEO-tool marketing content aren't fabricated, they're built from keyword sets weighted toward informational queries, the exact category of search where AI Overviews appear most. A tracker built primarily from "how to," "what is," and "best X for Y" queries will show a much higher AI Overview rate than one sampling a realistic mix of navigational, transactional, and informational search, because that's the category the feature was built for in the first place.
Neither approach is dishonest. A tool vendor selling AEO software has a real incentive to sample the queries where its product matters, and a broad tracker like Semrush's has an incentive to represent search as a whole. Read any single statistic with that in mind: what was the sample, and does it match the kind of query your business actually competes for?
What about ChatGPT and Perplexity specifically?
This is a separate question from AI Overviews, and the two get conflated constantly. Google's AI Overview is still a Google search result: the query starts and ends on google.com. ChatGPT and Perplexity are searches that never touch Google at all.
- Google still handles roughly 80-90% of total global search volume by most measures.
- Standalone AI platforms (ChatGPT, Perplexity, and similar tools) are estimated to be taking around 15-20% of informational and research-style query volume specifically, as of early 2026 — not 15-20% of search overall.
- That narrower slice, informational and research queries, is exactly the category where a business's blog, guides, and FAQ content compete, which is why it's the number worth watching even though it applies to a smaller share of total search.
Total search volume and informational-query volume are different denominators, and headlines routinely swap one for the other. A statement like "AI has taken 20% of search" and "AI platforms take 15-20% of informational query volume" sound almost identical and mean very different things.
What this actually changes about the work
Almost nothing, which is itself the finding worth sitting with. Whether the real AI Overview trigger rate for your industry's queries is 16% or 50%, the response looks the same either way: be crawlable, answer the actual question in a passage a model can lift whole, cover the sub-questions around your main topic, and don't build a business model around any single citation channel. See how to get your business cited by AI search for the specific mechanics.
The percentage is a headline. The underlying shift, that a meaningful and growing share of research-style queries never reach a list of ten blue links at all, is real regardless of which tracker's number you trust. Build for that shift, and the exact figure stops mattering very much.
If you want a straight read on where your own content actually stands, on Google and on AI platforms, we run that as part of a free audit.
Frequently asked questions
It depends entirely on which tracker you ask and what they're measuring. Semrush's own tracked panel of over 10 million keywords put AI Overview prevalence at about 15.7% by November 2025, after peaking near 24.6% in July. Other trackers that sample informational or long-tail-heavy keyword sets report figures from 43% up to 60% for roughly the same period. Neither is lying; they're measuring different query mixes.
Mostly sample composition. A tracker built from a broad, representative keyword set (transactional, navigational, and informational queries mixed) will show a lower AI Overview rate, because AI Overviews trigger far more often on informational "how" and "what is" queries than on searches like a brand name or a product model number. A tracker weighted toward informational keywords will report a much higher percentage for the exact same underlying reality.
Google still handles roughly 80-90% of total search volume by most measures. Where AI platforms have made a real dent is specifically informational and research-style queries, not search overall: estimates for that narrower slice put ChatGPT, Perplexity, and similar tools at around 15-20% of informational query volume as of early 2026. That is a meaningfully different claim than "AI has taken 20% of search."
Not much, and that's the useful conclusion. Whether the real AI Overview rate is 16% or 50%, the response is the same: be crawlable, answer questions in liftable passages, and don't depend on a single citation channel for leads. The exact percentage matters for headlines. It changes almost nothing about the work.