The short version: Google AI Overviews now appear on 43% of US searches, up from roughly 15% a year ago, according to Similarweb. That number is real, and it is almost certainly not your number. A separate study of 146 million search results found AI Overviews on just 2.1% of transactional queries and 7.9% of local ones. For most small businesses, the risk is not spread evenly across the site. It is concentrated in one specific place, and knowing where changes what you should do about it.
Two credible research outfits published wildly different figures about the same thing this month, and the gap between them is more useful than either number on its own.
Similarweb’s 2026 Generative AI Landscape report, covered by TechCrunch on July 27, found that Google’s AI-generated answers now show up on 43% of US searches, nearly triple the share from a year earlier. The same report tracked AI Mode, Google’s more conversational surface, growing from 126 million visits in June 2025 to 279 million by May 2026. Average query length is rising too, which suggests people are learning to type full questions instead of three-word fragments.
Read on its own, that reads like an emergency. Nearly half of all searches now answered before anyone clicks anything.
Then look at Ahrefs’ analysis of 146 million search results pages, which broke the same phenomenon down by what people were actually trying to do:
Informational queries: 21.4%. Commercial queries: 4.3%. Transactional queries: 2.1%. Navigational queries: 0.9%. Local searches: 7.9%, against 22.8% for non-local.
That is a tenfold spread between someone researching a topic and someone ready to hand over a credit card.
Why do the AI Overview numbers disagree so much?
Because they are answering different questions, and both answers are correct.
Ahrefs sampled keywords sitting in a database, on desktop, in September 2025. Similarweb measured searches that real humans actually ran, through May 2026. Those are different denominators. A keyword database treats “how do heat pumps work” and “heat pump repair Fresno” as one entry each. Real life does not. People run enormous volumes of curious, exploratory, question-shaped searches, and those are exactly the ones AI Overviews were built to answer. Add eight months of Google expanding coverage, and the two figures stop looking contradictory.
So the honest answer to “how exposed am I” is not 43%, and it is not 2.1%. It depends entirely on what kind of searches bring you customers.
Which of your pages are actually at risk?
Here is the part worth sitting with. The queries that make you money are, on this evidence, the least affected ones.
Somebody searching “emergency plumber near me” at 9pm is not being intercepted by an AI summary in most cases. Neither is somebody typing your business name, or searching to book, buy, or call. Those are navigational, local, and transactional queries, and they sit at the bottom of the trigger table. If your business runs on local intent and repeat customers, the 43% headline overstates your exposure by a wide margin.
What is genuinely exposed is the top of your funnel. The explainer posts. The “what does X cost” pages. The beginner guides you wrote to catch people early. That is where AI Overviews live, and if your traffic has softened this year, that is the first place to look. We covered the click-through side of this when ranking first started returning roughly half the clicks it used to.
The non-obvious read: an AI Overview is a signal, not just a loss
When Google puts an AI summary above your article, it is making a statement about that article. It is saying this question now has a settled, commodity answer that can be assembled from the open web in two seconds.
That is worth knowing. It means the piece was never much of a moat. Anyone could write it, and now a model can assemble it on demand.
Which points at what does survive. An AI Overview cannot tell someone what you charge. It cannot describe how your team handles a job, show the work you did on a street they recognise, name your warranty terms, or explain why you turned down a project like theirs last month. It cannot be your proof. Models summarise what is already public and general; they cannot manufacture what is specific and yours.
So the move is not to write more of what got summarised. It is to shift that effort toward the pages only you can write. That is a better use of the hours anyway, and it happens to be the content that converts.
How do I find my own AI Overview rate?
You can measure this yourself in about half an hour, with no new tools.
Open Google Search Console and pull your top 20 queries by impressions over the last three months. Next to each one, mark what the searcher wanted: to learn something, to compare options, or to buy or book. Then run those queries yourself, logged out, and note which ones return an AI Overview.
You now have your actual rate, and something more useful than a rate: a list. If the AI Overviews cluster on your “learn something” queries while your booking queries stay untouched, you are in a normal and manageable position. If they are showing up on your commercial queries, that is a real signal worth acting on quickly.
Check the summaries for accuracy while you are there. AI Overviews have misrepresented businesses before, sometimes badly. We wrote about the case where AI Overviews surfaced scammers’ phone numbers as official support lines.
What should a small business do about AI Overviews in 2026?
Three things, in order.
Protect the money queries. Your local listings, booking pages, and service-plus-city pages are the least disrupted and the most valuable. They deserve the attention that panic would otherwise send to your blog.
Stop competing on commodity answers. If a model can assemble it, you will not win it back by writing it again with better headings.
Write what only you can write. Prices, process, proof, and specifics. That lines up with the rest of our practical playbook for AI in small business marketing.
None of this needs new software or a bigger budget. It needs knowing which half of your site is actually under pressure, and that is a question you can now answer with your own data instead of somebody else’s average.
Frequently Asked Questions
Do AI Overviews appear on local searches?
Much less often than on general searches. Ahrefs found AI Overviews on 7.9% of local searches, compared with 22.8% of non-local ones. If your customers find you through “service plus city” or “near me” searches, your exposure is considerably lower than the widely quoted 43% figure suggests, though it is worth checking your own queries directly rather than assuming.
Is 43% or 21% the correct AI Overview figure?
Both, because they measure different things. Similarweb’s 43% reflects searches real people actually ran through May 2026, which skew heavily toward the question-shaped queries AI Overviews target. Ahrefs’ figures come from sampling keywords in a database on desktop in September 2025, a flatter and earlier sample. Neither is wrong, and neither describes your specific business.
Will an AI Overview stop people clicking my website?
On informational queries it frequently reduces clicks, because the searcher gets their answer without leaving the results page. On transactional and local queries the effect is much smaller, since someone who wants to book, buy, or call still needs to reach an actual business. The practical implication is that your traffic loss, if you have one, is probably concentrated in your educational content rather than spread across the whole site.
Should I stop writing blog content because of AI Overviews?
No, but the calculation has changed for one type of content. General explainer posts that any competitor could write are the most likely to be summarised and the least likely to earn a click. Content built on things only your business knows, your pricing, your process, your local knowledge, and your documented results, remains difficult for a model to replicate and tends to convert better anyway.
We are curious about one thing in particular: if you have checked your own search queries recently, did the AI Overviews land where you expected them to, or somewhere that surprised you?
