Impressions are steady, rankings look the same, and clicks have fallen. That pattern sends many site owners straight to AI Overviews, and sometimes they are right. Sometimes the cause is a ranking loss, a seasonal swing, a tracking change or a technical fault, and treating the wrong cause wastes the effort.
This guide puts the main studies side by side with their limits, gives you a way to tell an AI Overview loss from other causes using Google Search Console, and sets out what we do about it. We state what the data cannot tell you. If your question is how to get named inside the overview, the AI Overview optimization services page covers the service.
01Why AI Overviews reduce traffic
An AI Overview is a generated summary at the top of some Google results, with links to sources. When it answers the question, fewer people need to click anything. In a Pew Research Center study of 900 US adults and 68,879 Google searches, people clicked a traditional result in 8% of visits to pages with an AI summary and 15% of visits to pages without one. They clicked a link inside the summary in 1% of visits, and ended their browsing session in 26% of visits with a summary against 16% without one.
Ahrefs compared 300,000 keywords, half with an AI Overview and half informational keywords without one, using Search Console data for position one on desktop. It estimated the overview cut the click rate of the top result by about 58%. In the same data, informational keywords with no AI Overview also lost roughly half their position-one click rate, from 7.6% to 3.9%, which Ahrefs attributes to other result-page features and zero-click behaviour. So part of what looks like an overview effect is a wider trend.
The picture also moves. Seer Interactive's data on 53 brands saw organic click-through on AI Overview queries rise from 1.3% to 2.4% over two months after a low. Treat every figure here as a snapshot with a method attached, not as a constant.
02What the studies agree on, and where they stop
| Study | What it measured | Finding | Limit to keep in mind |
|---|---|---|---|
| Pew Research Center | Browsing of 900 US adults, 68,879 searches | Clicks on results 8% with a summary, 15% without | US adults only, Google only, a month of browsing |
| Ahrefs | 300,000 keywords, position one, desktop | About 58% lower click rate for the top result | Correlation; control group also declined |
| Seer Interactive | 53 brands, 5.47M queries | Click rate on AI Overview queries recovered from 1.3% to 2.4% in two months | Correlation; one account dominates some segments |
| Define Media Group, as reported by TSEG | 64 sites in Search Console | Clicks ended 42% below earlier benchmarks | Reported second-hand; no control for seasonality or algorithm changes |
Sources listed below the page
All four agree on the direction: fewer clicks where a summary appears. They disagree on size, because they measure different things in different places. The last one is useful as a warning and weak as evidence, since the report we read gave no detail on the sites and did not separate the overview from other causes.
03Which pages lose traffic first
Ahrefs found AI Overviews appear on about 20.5% of keywords and 57.9% of question queries. So question-style, informational pages are the exposed ones: what is, how to, definitions, simple comparisons. Pages that serve a decision, such as pricing, product and contact pages, are less exposed, because the summary cannot complete the purchase. Brand searches are the least exposed, because the searcher already wants you.
The bars matter because of what the middle and last rows say. Where an overview appears, a cited brand got more than twice the click rate of an uncited one in Seer's data. Seer's authors note this is correlation, higher-authority brands are also more likely to be cited, and one account made up a large share of the not-cited segment. The direction is still the point: being inside the overview is better than being absent from it.
04Diagnose the drop before you respond
Google says traffic from AI features is included in the overall Web totals in Search Console, with no separate line for it. So you cannot read the effect off one report. Instead, compare the same pages and queries over the same weeks the year before, then split the change by what happened to impressions, position and clicks.
| Impressions | Average position | Clicks | Likely cause | Next step |
|---|---|---|---|---|
| Steady or up | Steady | Down | AI Overview or other result-page features | Check the live result for your top queries |
| Down | Down | Down | Lost rankings, an update, or a technical fault | Check indexing, page errors and rivals |
| Down | Steady | Down | Less demand or seasonality | Compare with the same weeks a year earlier |
| Down on one section only | Steady | Down | Redirect, template or tracking change | Check that section's technical history |
| Steady | Steady | Down on question pages only | AI Overview on informational queries | Segment by query type |
Export queries and pages
In Search Console, export the last full period and the same period a year earlier, by query and by page.
Segment the queries
Split them into question and informational queries, brand queries and transactional queries. An AI Overview effect shows mostly in the first group.
Check the live result
Search your 20 biggest losing queries yourself. Note which show an overview, whether you are cited, and who is.
Rule out the other causes
Check indexing, errors, redirects, tracking and the algorithm update history before you blame the overview.
Put a value on the loss
Multiply the lost clicks by what a click is worth to you. Informational clicks are often worth far less than product and pricing clicks.
05What to do about an AI Overviews traffic drop
- Protect pages that earn revenue. Make sure pricing, product, comparison and contact pages are clear, current and indexed
- Get cited where an overview appears. Ahrefs found 37.9% of AI Overview citations come from Google's top 10, 31.2% from positions 11 to 100 and 31.0% from beyond 100, so rank is not the only way in. Clear answers and outside mentions count
- Give people a reason to click. A summary can state a definition. It cannot give a calculator, a template, original data or a worked example you can copy
- Keep pages concise when the answer is short. Ahrefs found 53.4% of cited pages in AI Overviews are under 1,000 words, so length is not what earns a citation
- Build the brand searches. Branded web mentions correlated 0.664 with AI Overview visibility in an Ahrefs study of 75,000 brands, so being talked about elsewhere pays twice
- Measure by outcome. Track leads and revenue from the pages that matter, not total clicks
06Which pages to rebuild first
Not every page that lost clicks deserves work. Score each losing page on two questions: does it lead to revenue, and does a summary fully replace it? A page that does both badly is a rebuild. A page that leads nowhere and is fully replaced may be better merged into a stronger page or left alone.
| Leads to revenue | Summary replaces it | Action |
|---|---|---|
| Yes | No | Protect: keep it clear, current and indexed |
| Yes | Yes | Rebuild: add what a summary cannot hold, such as data, a tool or a worked example |
| No | Yes | Merge into a stronger page, or accept the loss |
| No | No | Leave alone, and recheck after your next measurement |
Brand-new pages are held to the same rule. Before writing one, ask what a reader gets from it that an overview cannot give. If the honest answer is nothing, write something else. If you are choosing between Google and chat engines for the effort, GEO versus AEO sets out the difference, and engines outside Google each have their own rules, for example ChatGPT and Perplexity.
07Track whether you are cited
Search Console cannot tell you whether an overview cites you, so check by hand. Take your 20 most valuable informational queries, search each, and note whether an overview appears, which sources it links, and whether you are among them. Repeat monthly, because overviews change from one search to the next. The AI visibility checker builds the same kind of prompt set for chat engines. Keep a log labelled run one, run two and run three, so the comparison stays like for like. How we report this to clients is on the case studies page. If you find you are cited on some queries and not others, note what the cited pages have in common, such as a direct answer in the first lines, a table, or a named source, and copy that structure on the pages you lost. That comparison costs an hour and usually shows the next rebuild.
08The snippet control trap
Google documents nosnippet, data-nosnippet and max-snippet controls that limit how much of your page appears in results, and noindex removes the page entirely. It also says a page must be indexed and eligible to show with a snippet to be a supporting link in AI features. So restricting snippets can take you out of the overview altogether. Our decision rule: do not use these controls to protect traffic unless the content itself is the product you sell. For most sites, the better move is to be the page the overview cites and to offer something it cannot summarise.
09What nobody can tell you
People ask about a 30% rule for AI. We found no established rule behind that phrase, and the studies above show a wide range, so we would not plan around one number. People also ask whether Google is losing money on AI Overviews. We could not find a verified public figure, so we will not guess. And nobody can promise that a page will be cited. Our measurement guide shows how to track whether you are, prompt by prompt.
A one-page response plan
0 of 7 doneWhat most teams do after a traffic drop
- Blame AI Overviews without segmenting the data
- Rewrite every post that lost clicks
- Block snippets to protect content
- Report total clicks and nothing else
What we do
- Segment by query type and compare like periods
- Rebuild only the pages tied to revenue
- Keep pages eligible to be cited and add what a summary cannot
- Report citations in named prompts alongside leads and revenue
This guide suits you if
- Your clicks fell while impressions held
- Your traffic comes largely from question-style searches
- You can read Search Console, or have someone who can
Look elsewhere if
- Your rankings and impressions both collapsed, which points to a ranking or technical cause
- You want a guaranteed recovery, which no one can promise
If you want help, the starting point is an AI visibility audit, and our work is described under generative engine optimization services at $2,000 a month. See AI search statistics for the wider numbers, Google AI Mode optimization for the next Google surface, and the AI visibility hub for the whole map.
