AI Overviews have changed search behaviour more in the last two years than the previous fifteen did. People still search the same way — it’s what they do with the results that’s completely different. For any business that depends on being found, the consequences are only starting to show up in the numbers.
The click has quietly become optional
Think about the last thing you searched for. You typed the question, and before your eyes reached the first result, Google had already answered it — one of Google’s AI Overviews at the top of the page, assembled from sources you didn’t choose. You read it, got what you needed, and moved on.
That’s not a design change. It’s a behaviour change, and it’s measurable. The Pew Research Center tracked 68,879 real Google searches by 900 US adults, comparing pages with and without an AI summary:
| What happened | With AI summary | Without |
|---|---|---|
| Clicked a normal search result | 8% | 15% |
| Clicked a source link inside the summary | 1% | — |
| Ended the browsing session entirely | 26% | 16% |
Read that last row again. When AI Overviews answer the question, people don’t click a different result — they’re more likely to stop searching altogether.
Is ranking first still the top of the page?
Not really. For twenty years, SEO has been a competition for the first organic result, because that’s where the attention was. That assumption no longer holds. The first result is still there — but now it sits underneath an AI Overview that has already answered the question.
Ranking first and being read are no longer the same thing. You can hold position one, watch impressions stay flat or even rise, and still see clicks fall. Plenty of site owners have spent a year trying to work out what they did wrong, when the honest answer is: nothing changed on their site. What changed is what sits above it.
The question is no longer only how do I rank. It’s also: what do AI Overviews say about you, and are they right?
And the answer isn’t always correct
This is where it gets uncomfortable. The New York Times, working with the AI company Oumi, ran 4,326 searches through SimpleQA — a standard accuracy benchmark — and found Google’s AI Overviews accurate around 91% of the time.
Ninety-one percent sounds reassuring until you apply it to Google’s scale: across trillions of searches a year, that’s tens of millions of wrong answers every hour.
The finding that matters more for businesses is different. Over half of the answers that were factually correct were still ungrounded — the sources Google linked to didn’t actually support what the summary said. The answer happened to be right; the citation didn’t back it up. (Google disputes the analysis, arguing the benchmark has holes and that AI Overviews already warn they can make mistakes — fair, but it doesn’t change the reality for a brand owner.)
Three ways AI Overviews go wrong for a brand
- You’re not in the answer at all. A competitor is named, you aren’t, and the searcher never scrolls far enough to find you. The one most people worry about, and the easiest to spot.
- You’re in the answer, but out of date. Old pricing, a discontinued product, a name you traded under three years ago. The model pulls from whatever it can find, and the internet has a long memory. More common than most companies realise, because nobody’s checking.
- You’re cited for something you never said. The ungrounded problem — your page listed under a claim it doesn’t make. To the reader, that carries your authority. If it’s wrong, or about a competitor, that’s still your name on it.
What you can actually do about AI Overviews
You can’t edit AI Overviews directly. You can change what they have to work with — and in practice that comes down to a handful of things:
- Find out what it currently says. Search your brand, your main products and your customers’ real questions, and read the answers properly. Most companies have never done this once.
- Make the facts easy to extract. Models lift clear statements, not marketing copy. If the key facts are buried inside a paragraph about passion and innovation, they won’t be picked up. Say the thing plainly, somewhere obvious.
- Fix what’s out of date on your own site first. Old pricing, stale service descriptions, superseded product names — usually where the wrong information starts.
- Answer the actual questions. Clear headings, direct answers underneath, proper structured data. A page that answers a question in its first two sentences is far more likely to be the source that gets used.
- Get accurate mentions on the sources models trust. Pew found the most-cited sources were Wikipedia, YouTube and Reddit. Your own site is one input among many; what third-party sites say about you carries real weight.
None of this is a trick, and none of it is fast. It’s the same principle that always applied to search: accurate, clearly written, well-cited information wins. What’s changed is the cost of getting it wrong. A bad ranking used to mean fewer visitors. Bad information in AI Overviews means the wrong thing is being told to people who’ll never visit at all.
Where AWISEE fits
We work on AI mentions and AI visibility, which in practice means two things. We go through your content and fix it so the information AI Overviews pull about you is accurate, current and written in a way models can use. And we build accurate mentions of your brand on the third-party sites these systems actually reference — so the picture being assembled about you is one you’d recognise. It’s the core of our AI visibility services.
If you want to know what AI Overviews are currently saying about your brand, that’s a sensible place to start. Get in touch and we’ll take a look.
Sources
- Pew Research Center — Google users are less likely to click on links when an AI summary appears in the results.
- Search Engine Land — Google AI Overviews accuracy analysis.