Rankings Steady, Traffic Down? Look for the Culprit in AI Overviews
A familiar picture: rankings hold steady, impressions in Google Search Console stay put, and traffic quietly slides down. Your first thought — "Google tweaked the algorithm again." But in reality, the culprit is often sitting right above your snippet. It's AI Overview.
Is this the new normal? Yes. This is the new reality of search, and you need to work with it as soberly as with any other ranking factor.
Why your rankings are lying to you
Reading GSC metrics used to be simple: position goes up, traffic goes up. Now there's a layer in between. Google generates an AI answer right in the results, the user gets the gist of their question without scrolling to organic listings, and they don't click at all. You're still in third place, you're still being shown — but there's no click. Position and impressions are facts of being displayed. CTR is a fact of interest. And it's CTR that AI Overview cuts first, while the other metrics stay untouched and create an illusion that everything's fine.
Here's the core of the problem in one sentence: looking at rankings separately from CTR is no longer enough. You need to split these metrics and watch CTR as a signal in its own right, not as a byproduct of your position.
How to check this in practice
Pull up the queries report in GSC and sort by CTR drop where position is stable or growing. That's your suspect list. Then go manual — open the search results for each of those queries and check whether AI Overview is there. If it is, and your site is even cited as a source inside the block — congratulations, you're feeding someone else's answer instead of your own traffic.
Pay attention to the nature of the query itself, too. Informational queries — "what is," "how," "why" — are AI Overview's bread and butter, they're practically begging for it. Transactional and narrow niche queries get hit less — the model is less confident there, and Google is more willing to send the user further down the links.
What to do about it
Give first, take later — the old rule works here too, just from a new angle. If your content is exactly the same fact that AI Overview pulls out, you're handing value to the model for free and getting nothing back. Which means your content needs a reason for someone to keep scrolling after they've read the summary.
The first path is to become a source for AI Overview itself. Clear facts, numbers, definitions, structured data — the model picks up material more readily where meaning is easy to extract. This doesn't guarantee a click, but it raises your odds of being noticed and cited, and some of that audience will still go check the original source.
The second path is the opposite direction. Don't try to be an exhaustive short answer — give what can't be compressed into a short answer in the first place: a personal breakdown, an opinion, a case study, a calculation tailored to the reader's specific situation. AI Overview is good at facts and bad at position and experience. This is exactly where a writer with real practice in the field has an edge no model can copy.
And third, the most uncomfortable one for a lot of people: stop praying to a single traffic source. If your content is clearly informational and vulnerable to AI Overview, build backup channels in early — a newsletter, social media, direct visits from a loyal audience. A captain doesn't know where he's headed if he's got only one sail, and that sail is Google organic.
Bottom line
AI Overview isn't the apocalypse, and it's not a reason to drop SEO. But it's a new filter between you and the user, one you need to account for when planning content from the start — not something you dissect after the fact, once traffic has already tanked. Split the metrics in GSC, find your vulnerable queries, and decide for each one: play the AI-source game, or build content AI can't replace. If you're not sure where to start untangling your own situation, we can talk it through in a consultation via the services page.