How to Recover From a Traffic Drop Caused by AI Overviews
First confirm that AI Overviews caused the drop by comparing question-style queries with the rest in Search Console. Then move your effort to queries an AI summary cannot fully answer, make your page the source the summary cites, and add routes to buyers that do not depend on one click. Expect a smaller click total on question-style queries; whether the remaining visitors convert better is something to measure on your own site 25.
- Pew found people clicked a normal result in 8% of visits with an AI summary and 15% without one 2.
- AI Overview clicks and impressions sit inside your normal Search Console totals, so you must isolate them yourself 1.
- Studies disagree on the size of the loss, and Google says total clicks are stable 35. Measure your own site.
- Brands cited inside an AI Overview earned a higher click rate than uncited brands in one study 4.
- Do not rewrite everything. Fix the pages that lost clicks on queries where a summary now answers the question.
What do you need to diagnose the drop?
You need read access to Google Search Console, a spreadsheet, and about three hours for the diagnosis. The fixes then take four to eight weeks of normal content work, and Google advises waiting a few weeks before judging a change 9.
You do not need a paid tool. A paid AI-visibility tracker helps later, but the first job is to find out which of your pages lost clicks and why. Search Console already holds that answer.
Was it really AI Overviews that cut your traffic?
Not always. A drop can come from a ranking change, a seasonal dip, a broken page, a core update or a Google answer box that existed before AI Overviews. Treat AI Overviews as a suspect until your own data points to it.
Google makes this harder. It counts clicks and impressions from AI features inside the normal Web search totals in Search Console, with no switch to filter them out 1. A click on an external link inside an AI Overview counts as a click, and the whole Overview takes one position, so every link in it shares that position 6.
That means you must infer the cause from patterns. The strongest clue is a drop in clicks with stable or rising impressions on informational, question-style queries. Your page is still being shown, but fewer people need to open it.
How big is the click loss, honestly?
Published numbers range from modest to severe, and they measure different things. Pew Research tracked 900 US adults in March 2025. People clicked a traditional result in 8% of visits when an AI summary appeared and 15% when it did not, and clicked a link inside the summary in 1% of visits 2. Pew says this shows an association, not proof that summaries cause the gap 2.
Ahrefs compared 300,000 keywords and found that an AI Overview went with a 58% lower average desktop click rate for the position-one page, between December 2023 and December 2025. It labels the result as correlation and builds its no-Overview baseline from a model 3.
Google's view is different. Its head of Search wrote that total organic clicks to websites were relatively stable year over year, and that more clicks are quality clicks, meaning users do not quickly click back 5. Google published no dataset with that claim. The honest reading is that averages hide big differences between sites and query types, so your own numbers matter more than any study.
| Source | What it measured | Finding | Main limit |
|---|---|---|---|
| Pew Research 2 | Clicks per Google results page visit, 900 US adults, March 2025 | 8% click a result with a summary, 15% without | One month, US only, association only |
| Ahrefs 3 | Average desktop click rate for the position-one page, 300,000 keywords, 2023 vs 2025 | 58% lower with an AI Overview | Modelled baseline, correlation |
| Google 5 | Total organic clicks to websites | Relatively stable year over year | No data published |
How do you find which pages and queries lost clicks?
Export two equal date ranges from the Performance report: the 90 days before the drop and the 90 days after. Compare clicks, impressions and click-through rate by page and by query. Sort by the largest click loss.
Then tag each lost query by type. Question-style and long queries are the ones most likely to show a summary. Pew found summaries on 8% of one- or two-word searches, but on 53% of searches with ten or more words and 60% of searches starting with a question word such as who, what or why 2.
Group the losers into three buckets: pages whose impressions held but clicks fell, pages where both fell, and pages that are fine. Only the first bucket points at AI Overviews. Google also lists a Generative AI performance report in Search Console as the place to measure how you do in these features 7.
- Impressions steady, clicks down: likely an AI summary or other answer box.
- Impressions and clicks both down: look at rankings, indexing and seasonality first.
- Clicks steady or up: leave the page alone.
Which queries should you stop chasing, and which should you move toward?
Stop trying to win clicks on pure definition queries, such as what a term means. A summary answers those in one paragraph. Move toward queries where the reader needs more than a paragraph: comparisons, calculators, templates, local detail, current prices and decisions with trade-offs.
Google's own advice points the same way. It calls non-commodity content, meaning a unique point of view based on first-hand experience instead of recycled summaries, the most influential factor in the long run 7. A page that repeats what ten others say is the page an AI summary replaces.
- Page: 'What is a retainer agreement?' Impressions steady, clicks down. Decision: keep as a short definition, stop investing.
- Page: 'Retainer agreement template for a design studio'. Clicks steady. Decision: expand with a downloadable template and a filled-in sample.
- Page: 'Retainer vs project fee: which costs less?' Clicks falling. Decision: add a calculator and a decision table the summary cannot reproduce.
- Result of the review: 1 page left alone, 1 page expanded, 1 page rebuilt around a tool.
How do you become the page the AI Overview cites?
To be a supporting link, a page must be indexed and eligible to show in Google Search with a snippet. Google says no extra technical requirements exist for AI Overviews and AI Mode 1. So the work is content quality and clarity, not a special file.
Being cited seems to pay. In Seer Interactive's data on 3,119 informational terms, brands cited in an Overview earned a 0.70% organic click rate against 0.52% for brands not cited. Seer warns that stronger brands may simply get cited more, so this is not proof of cause 4.
Citations also come from a wider set of pages than the classic top ten. Ahrefs found that about 38% of cited URLs also appeared in the top ten results in a study published in March 2026, against about 76% in July 2025. It suggests, but does not prove, that query fan-out explains the shift: Google splits a question into sub-queries and cites pages that appear across them 8. Cover the sub-questions around your topic on one useful page. Do not split them into thin pages, because Google warns that pages made only to match every variation can break its scaled content abuse policy 7.
How do you reduce your reliance on one Google click?
Build ways to reach buyers that skip the results page. Ask visitors to join an email list on your best pages, because a subscriber is not hostage to a summary. Publish where your buyers already look, such as YouTube, LinkedIn and industry communities, and encourage branded searches by making your name easy to remember.
Also earn mentions on other sites. AI answers draw on pages beyond your own, so a clear presence on review sites and in trade press gives you more than one way into an answer. See how to earn brand mentions with digital PR and how to build topical authority.
What mistakes make the drop worse?
The common errors are rewriting every page, deleting pages that still bring leads, and chasing every sub-query with a new thin page. Another is blocking Google from using your content with a snippet control to protect traffic. Google says those controls limit what shows in AI features 1, and a page with no eligible snippet cannot be a supporting link.
A quieter mistake is judging by sessions alone. If a page loses clicks but the remaining visitors enquire at the same rate, the business effect may be small. Track enquiries next to clicks.
How do you check the recovery is working?
Pick a fixed set of 10 to 20 pages and 20 prompts before you change anything. Record clicks, impressions, enquiries and whether your pages appear in the Overview. Re-check at four and eight weeks, and compare with pages you did not touch.
Recovery here means a better mix, not always the old total. A smaller audience that is better qualified, with fewer clicks turning into more enquiries, is a possible result worth checking for. Our guide to testing whether your AI SEO changes worked covers the design in more detail.
When should you not act?
Do not act on one week of data, on queries that never brought leads, or on a drop that matches last year's seasonal dip. Do not rebuild a site because of a headline number from a study about other sites.
If clicks fell but leads held, you have a reporting problem, not a traffic problem. Fix the report first, then decide whether anything needs changing.
Frequently asked questions
Can I see AI Overview traffic separately in Search Console?
Not directly. Google counts clicks and impressions from AI features inside the normal Web search totals 1. You can infer the effect by filtering question-style queries and comparing clicks with impressions, and use the Generative AI performance report Google points to for your presence in these features 7.
Will blocking Google from showing snippets bring my clicks back?
No. Snippet controls limit what Google shows from your page, including in AI features 1. A page that cannot show a snippet cannot be a supporting link, so you would lose both the citation and the visibility. It is rarely the right trade.
Is a cited page better than a page ranking first?
Not always, and the data is mixed. Seer saw higher click rates for cited brands but warns stronger brands may be cited more often 4. Aim to rank and to be cited. Treat the citation as one more way to be seen, not as a replacement for ranking.
Should I delete pages that lost traffic?
Usually not. A page that lost clicks may still bring enquiries or support other pages. Check enquiries and internal links first. Rebuild it around original material, merge it into a stronger page, or leave it as a short definition, as the triage table suggests.
How long until I know if the fixes worked?
Plan on four to eight weeks of data, since Google says results can take anywhere from hours to months to appear 9. Compare changed pages with untouched control pages, since the whole market moves at the same time.
Sources
- AI features and your website · Google Search CentralOpened October 20, 2026. Supports: AI feature traffic is reported inside the Performance report under Web search; a page must be indexed and eligible to show with a snippet; no extra requirements; snippet controls limit what shows. Page last updated 2025-12-10.
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research CenterOpened October 20, 2026. Supports: 900 US adults, March 2025: 8% vs 15% click on a traditional result with and without a summary; 1% click inside the summary; summaries appeared on 8% of 1-2 word searches, 53% of searches of 10+ words, 60% of question-word searches; association not proof of cause.
- AI Overviews Reduce Clicks by 58% · AhrefsOpened October 20, 2026. Supports: 300,000 keywords; AI Overview correlates with a 58% lower average click-through rate for the top-ranking page (Dec 2023 vs Dec 2025); modelled baseline; correlation not causation. Published 4 February 2026.
- AIO Impact on Google CTR: September 2025 Update · Seer InteractiveOpened October 20, 2026. Supports: 3,119 informational terms, 42 organisations: cited brands 0.70% organic CTR vs 0.52% uncited on AI Overview queries; author warns that stronger brands may be cited more often.
- AI in Search is driving more queries and higher quality clicks · Google (Liz Reid)Opened October 20, 2026. Supports: Google says total organic clicks are relatively stable year over year and that quality clicks, where users do not quickly click back, are up. No dataset published. 6 August 2025.
- What are impressions, position, and clicks? · Google Search Console HelpOpened October 20, 2026. Supports: A click on an external link in an AI Overview counts as a click; standard impression rules apply; an AI Overview takes one position and all links in it share it.
- Optimizing your website for generative AI features on Google Search · Google Search CentralOpened October 20, 2026. Supports: Non-commodity content with a unique point of view is the most influential long-run factor; use the Generative AI performance report in Search Console; separate content for every fan-out variation to manipulate results can breach the scaled content abuse policy. Last updated 2026-07-10.
- Update: 38% of AI Overview Citations Pull From The Top 10 · AhrefsOpened October 20, 2026. Supports: 37.9% of cited URLs also in the top 10 (863K keyword SERPs, 4M URLs, March 2026) vs about 76% in July 2025; Ahrefs attributes the shift to query fan-out.
- SEO Starter Guide: The Basics · Google Search CentralOpened October 20, 2026. Supports: Changes can take a few hours to several months to show; wait a few weeks to assess whether work had beneficial effects. Last updated 2025-12-10.