AI Overviews optimization
Your rankings held and your clicks fell off a cliff. That pattern has a name, a measured size, and three sane responses - none of which is writing more of the content that got eaten.
What AI Overviews are
An AI Overview is the generated answer Google places above the organic results, assembled from multiple sources and carrying links to some of them. For the searcher it is a summary that often ends the search. For you it is a competitor that sits above every position you've ever held and was built partly from your own pages.
They don't appear on everything. Triggering skews toward informational, question-shaped and how-to queries, and away from navigational and clearly transactional ones. Which of your queries trigger an overview is not something you can look up - you have to go and check them, one at a time, and that check is the first honest piece of work on this page.
What the damage actually measures
Two numbers are worth committing to memory, both from published research rather than vendor marketing.
| Finding | Figure | Source |
|---|---|---|
| CTR loss, top organic result, when an AI Overview is present | ~58% reduction | Ahrefs study, December 2025 |
| Share of searches ending without a click | Rose from ~56% to 69% in a year | Ahrefs study, December 2025 |
Read those together and the picture is unambiguous: the top spot is worth roughly half what it used to be on any query that triggers an overview, and the share of queries that end in no click at all is now the majority. If your plan for 2026 is "rank higher", you are competing harder for a prize that has been quietly devalued.
Two caveats, because the numbers get quoted carelessly. First, the effect concentrates on informational queries; a page that ranks for "buy X" or your brand name is far less exposed than one ranking for "what is X". Second, an aggregate is not a forecast - your own drop could be smaller or considerably worse depending on your query mix. Use the studies to size the risk, then measure your own site.
Was it AI Overviews or a core update?
These two get confused constantly and the fixes are opposite. A core update means Google changed its mind about your quality; an AI Overview means Google still likes you fine and gave your answer away anyway. Diagnose before you rewrite anything.
| What you see | Points to AI Overviews | Points to a ranking change |
|---|---|---|
| Impressions vs clicks | Impressions flat or rising, clicks falling | Both falling together |
| Average position | Roughly unchanged | Visibly worse |
| Which queries | Informational and question-shaped ones first | Broad, including brand and commercial terms |
| Shape of the decline | Gradual, query by query, over weeks | Sharp, dated, aligned to an announced rollout |
| Manual SERP check | The query now shows an overview | No overview; you've simply moved down |
The impressions-versus-clicks divergence is the single most useful signal you have, because Search Console does not break out AI Overview appearances as their own dimension. You are inferring the cause from the shape of the data, which is imperfect but perfectly serviceable: positions held, impressions held, clicks gone is a fingerprint, and it isn't the fingerprint of a core update.
Response one - be the cited source
If the overview is going to appear regardless, the remaining prize is being one of the sources it links. That traffic is smaller than the old blue-link traffic and it is not nothing, and the citation itself carries a signal to the reader that you are the authority on the thing they just asked.
The work is the same work that gets you cited by any assistant, because the underlying problem is identical: a synthesis layer looking for extractable passages. Answer one question per page, put the answer in the first self-contained paragraph beneath the heading, keep the content in raw HTML, date it visibly, and give it something specific to quote. The mechanics are broken down here.
What does not work is trying to hide. The nosnippet and max-snippet directives restrict how your content can be shown, but they restrict your ordinary search snippets at the same time, which for most sites costs more than it saves. Opting out of being summarised generally means opting out of being seen.
Response two - shift your query mix
The more durable response is to stop competing where a paragraph can replace you. This is a content strategy decision, not a technical one, and it's the one most people avoid because it means admitting some of their existing pages are no longer worth maintaining.
- Comparisons and situational answers. "X vs Y for a two-person team" depends on the reader's circumstances, which is exactly what a one-paragraph summary handles badly.
- Original data. If the fact exists only on your site, the overview must cite you to state it. This is the cleanest defence available and almost nobody uses it.
- Tools, calculators and anything interactive. A summary cannot perform a calculation for the reader; it can only tell them where one lives.
- Opinion and judgement. Synthesis flattens toward consensus. A defensible, specific position is the thing consensus cannot produce.
- Transactional and navigational intent. Less exposed by nature, because the searcher still has to arrive somewhere to finish the job.
The corollary is uncomfortable and worth saying plainly: a page whose entire value is restating information available in twenty other places is now a page with no future. That was arguably always true. The overview just made the bill arrive.
Response three - measure the divergence, monthly
- Track clicks-per-impression by query group, not just total traffic. Split informational queries from commercial ones and watch them separately; the aggregate hides the whole story.
- Flag any query where impressions rose and clicks fell. That list is your exposure register, and it should be reviewed monthly rather than discovered during a bad quarter.
- Spot-check the top twenty affected queries by hand to confirm an overview is actually present. Five minutes of looking beats an afternoon of theorising.
- Segment assistant referral traffic separately. Small volume, high intent, invisible if you leave it merged into direct.
- Record the dates. Without a dated baseline you cannot tell a platform change from your own change, and you will attribute both to whichever you were working on at the time.
This site is a new domain publishing its own numbers as they come in, unflattering ones included - across four sites launched in 2026, first-weeks impressions ran 2,200 to 16,500, click-through rates 0.1% to 0.6%, average positions 12 to 49. Predictions are logged with dates before the outcome is known and graded when they come due, on the receipts page.
An honest note
Nobody outside Google can see what makes an overview trigger, which sources it selects, or how that selection is weighted. The studies cited above measure outcomes, not mechanisms, and they are aggregates across many sites rather than a model of yours. Any tactic presented as a way to "get into AI Overviews" reliably is inference wearing a confident hat.
What survives that uncertainty is the boring part: be retrievable, answer one question per page, publish things only you can publish, and measure the divergence between impressions and clicks so you find out early rather than at the end of a quarter. All of it pays off in classic search too, which is the sensible reason to do it now - the wider case is on the GEO pillar, and if you'd rather have it audited than guessed at, that's what the audit is for.
Want to know which of your pages are being answered without you?
The GEO Audit maps your category's real buying prompts across AI Overviews and the major assistants, identifies where the answer is being given without a click, and returns the specific fixes - with a 90-day recheck so you find out whether they worked.