Why traffic drops even though rankings hold: AI Overviews explained to the client
It's the paradox breaking the most reporting meetings this year: the rankings report says everything is fine — the client is still first or second for their usual keywords — and the Analytics one says organic traffic has been falling for six months. The client looks at the two charts, can't understand how both can be true at once, and reaches the logical conclusion from where they're sitting: "you must be doing something wrong".
You're not doing anything wrong. But if you can't explain the mechanism clearly, you'll be the one paying a bill Google generated. This article is the full explanation: what's happening, how to prove it with the client's own data, and how to reframe the conversation before the reporting turns into a trial.
The mechanism: the answer eats the click
For twenty years, the deal with Google was stable: you ranked, Google showed your link, the user clicked. Position was a reliable proxy for traffic, which is why ranking reports worked as currency.
AI Overviews — those AI-generated summaries Google places above the results — break the deal without touching the rankings. Your client is still in the same position; what's changed is what's above them: a ready-cooked answer that resolves the user's question without them needing to visit any website. The ranking holds; the click vanishes.
The numbers behind the phenomenon, with their sources:
- Where AI Overviews appear, average organic CTR drops 61% in Spain (ismajimenez.com). Same spot, less than half the clicks.
- Zero-click searches — the ones that end without a click to any result — have gone from 56% in May 2024 to 69% in May 2025 (data cited on stucom.com). Seven out of ten searches now produce no visit at all.
- And the trend hasn't peaked: Gartner forecasts 25% less traditional search volume before the end of 2026 (press release, February 2024).
The analogy that works best with clients: their website was a shop on the high street, and ranking was having the best window display. Google has put a kiosk in front of the window that answers passers-by before they ever step in. The shop is still in the same spot — that's what the rankings say — but fewer people come through the door.
How to prove it in Search Console (the proof from their own data)
The advantage of this explanation is that it doesn't require the client to take your word for it: it can be seen in their own data in five minutes. The footprint of AI Overviews in Search Console is unmistakable:
- Open Performance → Search results and select the last 16 months.
- Switch on all four metrics: clicks, impressions, CTR and average position.
- Look for the scissor pattern: impressions stable or even rising, average position stable, and clicks and CTR falling. The impressions tell you Google is still showing the client as much as ever; the CTR tells you the user clicks less and less. If the position holds and the CTR falls, the explanation isn't in your work: it's in what Google has placed above.
- Filter by informational queries ("what is", "how", "which is best", prices, comparisons): that's where AI Overviews appear most and where the scissors open the widest. For branded or very transactional searches, CTR holds up better, and that contrast within the same account is the cleanest proof that the problem is the type of result, not the site.
Bring that chart to the meeting before the client brings theirs. The difference between "traffic has dropped" (said by them) and "traffic has dropped, here's the exact cause, and this is what we're going to do" (said by you, with their Search Console on screen) is the difference between defending the retainer and directing the strategy. How to build that whole meeting, chart by chart, is in what to show in the reporting meeting.
The reframe: from selling visits to selling presence
Here comes the uncomfortable part: explaining the mechanism clears you today, but it doesn't save you tomorrow. If your value proposition is still "traffic", you're selling a metric the market has decided to shrink year after year, and no explanation stops that erosion.
The honest reframe has three pieces:
First: the surviving click is worth more. Traffic arriving from AI answers converts at 14.2%, against 2.8% for classic organic — five times more (roymo.es, citing industry data). Whoever clicks after reading an AI answer arrives filtered and half-convinced. Fewer visits doesn't automatically mean fewer clients; the report should look at conversions before sessions.
Second: the new battle is appearing in the answer, not below it. If 69% of searches end without a click, the relevant question is no longer just "what position are you in?" but "what does the answer say about you?". That presence can be measured and worked on, and it should enter the monthly report alongside the rankings. Which specific metrics to add without promising what you don't control is in realistic AI visibility goals.
Third: the same phenomenon, in ChatGPT, doesn't even leave impressions. AI Overviews are the visible version of the problem because they happen inside Google and leave a trail in Search Console. Conversations in ChatGPT, Gemini or Perplexity leave none: if the AI recommends your client's competitor, there's no console where you can see it. For that you have to measure directly, asking the AIs what their audience would ask. We solve it with Surfeo, which fires those questions every week at the 4 AIs and leaves the result ready for the client's report, but the idea matters more than the tool: what isn't measured, in this channel, simply doesn't exist.
And there's one particular case that deserves separate watching: when what's falling isn't generic traffic but searches for the brand itself. That pattern has a different cause and a different treatment, which we develop in your branded search is falling: how to tell if AI is answering for you.
Frequently asked questions
How do I know whether a specific keyword shows an AI Overview?
Search it on Google in an incognito window and look. To do it at scale, filter in Search Console for queries with the biggest CTR drop while holding position: it's the most reliable indirect indicator. Informational and comparison queries are the near-certain candidates; navigational and very transactional ones, the most resistant.
Do I tell the client SEO no longer works?
No, because it's false: the AIs build their answers from content that ranks and from sources with authority, so SEO is the raw material of AI visibility. What no longer works is measuring success only in sessions. The work continues; the scoreboard changes.
And if the client asks why I didn't warn them sooner?
Honesty: this change has accelerated over the last two years and this conversation is the warning. Arriving with the mechanism explained, the proof in their Search Console and a plan is warning them in time. The alternative — waiting for them to find out from a headline or from another agency — would be arriving late.
Can the stable-impressions + falling-clicks pattern have another cause?
It can: a botched migration, a penalty, a new competitor dominating the SERP, or seasonality also move CTR. That's why a serious diagnosis cross-checks position (stable?), query type (does it fall more on informational?) and dates (does it coincide with the AI Overviews rollout?). If all three point to the same place, you've got your culprit.
Half of this conversation is won before the meeting: take the free AI visibility test with your client's website and arrive with the data their Search Console can't give them.
