Insight · 14 min
Rankings held, clicks fell
When Search Console shows stable positions and impressions but falling clicks, AI Overviews are the most likely cause for informational queries, and an unlikely one for branded and transactional queries. Every major study finds lower click-through rates where an AI Overview appears, but they disagree on the size, from about 15% to about 60%, because they measure different things. As of November 2026, Search Console does not tell you which of your queries show an AI Overview, so the diagnosis is a comparison you build yourself: query groups with and without an overview, same positions, same weeks of the year.
In short
- The studies agree on direction and disagree on size. Ahrefs estimates position-one CTR is about 58% lower than forecast where an AI Overview appears. Amsive measured an average 15.49% drop across 10 sites. Pew found people clicked a traditional result on 8% of visits with an AI summary and 15% without.
- Search Console counts AI Overview and AI Mode traffic inside the “Web” search type and does not flag which queries showed one. The Generative AI performance report shows impressions by page, not by query, and no clicks.
- The test is a comparison: tag each query by intent and by whether an overview appears, then compare CTR change within the same position band and the same weeks of the year. If only the overview group fell, the overview is the likely cause.
- Rule out the other causes first: seasonality, new SERP features, average-position artifacts, the September 2025 change in impression counting, and rewritten title links.
- Definitional and how-to pages lose most. Branded, pricing, product, and comparison pages lose least. Move effort toward queries that still click, compete for the citation on the rest, and report exposure as its own number.
What did the CTR studies find?
They found a drop every time, with sizes that range by a factor of four. The table puts the four most-quoted studies side by side.
| Study | Sample and method | Headline finding |
|---|---|---|
| Ahrefs, Feb 4, 2026 | 300,000 keywords, half with an AI Overview and half informational without one. Aggregated desktop Search Console data, December 2023 vs December 2025. | Position-one CTR on overview keywords was 0.016 against a forecast of 0.037 without the overview, an estimated 58% reduction. |
| Seer Interactive, Nov 4, 2025, as reported by PPC Land | 3,119 informational queries across 42 client organizations, June 2024 to September 2025, from Search Console and Google Ads. | Organic CTR on overview queries fell from 1.76% to 0.61% (61%). Queries without an overview also fell, by about 41%. |
| Amsive, Apr 16, 2025 | 700,000 keywords across 10 sites in 5 industries, focused on 10,000 that triggered an overview and already ranked. | CTR fell 15.49% on average. Non-branded fell 19.98%. Branded keywords that showed an overview rose 18.68%. |
| Pew Research Center, Jul 22, 2025 | Browsing data from 900 U.S. adults in March 2025: 68,879 Google searches, 12,593 with an AI summary. | Users clicked a traditional result on 8% of visits with a summary and 15% without. They clicked a link inside the summary on about 1%. |
Why don’t the CTR studies agree?
They disagree because each one answers a different question with a different sample.
- What is being compared. Ahrefs compares actual CTR against a forecast of what CTR would have been without the overview, so its 58% is an estimate built on a model. Seer compares the same queries across 15 months, so its 61% includes everything else that changed in that time, which is why its non-overview queries also fell 41%. Pew compares visits with and without a summary, which is behavior, not CTR on your site.
- Whose data. Ahrefs aggregates Search Console data across many sites. Seer uses 42 of its own clients. Amsive uses 10 sites. Pew uses a panel of people. A small set of sites can be dominated by one industry.
- Which queries. Seer covers informational queries only. Amsive includes branded terms, and branded terms that showed an overview gained clicks. Mixing intents pulls the average toward the middle.
- When. The studies span March 2025 to December 2025. AI Overviews changed coverage in that time, and Ahrefs notes that Gemini 3 has powered AI Overviews since January 2026, after the data for all four studies was collected.
- How CTR is computed. Amsive does not say whether its percentages are relative or point changes, or which periods it compared. Ahrefs uses desktop only.
The useful reading: on informational queries where an overview appears, expect a meaningful CTR loss, and do not assume your number matches any study’s. Measure your own. The AI search statistics reference lists these and other studies with what each one counted.
What does Search Console show for AI Overviews and AI Mode?
As of November 2026, Search Console counts AI Overview and AI Mode activity inside the standard Performance report and does not separate it. Google’s help page on impressions, position, and clicks sets the rules:
- A click on an external link in an AI Overview or in AI Mode counts as a click. Standard impression rules apply.
- An AI Overview occupies a single position, and every link inside it is assigned that same position.
- Position in AI Mode follows the same method as a normal results page.
- The traffic sits inside the “Web” search type, with no filter for AI features.
The Generative AI performance report, announced on June 3, 2026 and rolled out to all sites by August 31, 2026, adds one view. It shows impressions of your links in AI Overviews and AI Mode, by page, country, device, and date. It does not show queries, clicks, or position, and it reports the two surfaces together.
So Search Console can tell you which pages appear in AI features and how often. It cannot tell you which queries triggered an overview, and it cannot give you a CTR for overview results. That is the gap the diagnosis has to fill.
How do you find which of your queries show an AI Overview?
You find out from outside Search Console, with one of three sources, and you should note which one you used.
- A rank tracker that records SERP features. Most commercial trackers now flag whether a tracked keyword showed an AI Overview on the day it was checked. This is the fastest route for a few hundred queries. It is one location, one device, one check, so treat it as a sample.
- A manual check of your top queries. Search each of your top 50 to 100 queries by clicks lost, signed out, in the country you sell to, and log whether an overview appeared and whether you were cited. AI Overviews appear only when Google judges them useful, and the same query can show one on one day and not the next. Repeat the check on a second day for anything borderline.
- The Generative AI performance report, by page. Pages with many AI feature impressions are pages whose queries show overviews. Match those pages to their queries in the standard report. This tells you where to look, not which query did it.
Record the result as a column on your query export: overview present, overview absent, or unknown. The measurement guide covers the logging method in more detail.
How do you test whether AI Overviews caused the drop?
Compare CTR change between query groups that differ only in whether an overview appears. If the overview group fell and the matched group did not, the overview is the likely cause.
- Export two comparable periods. Use the same weeks of the year: this year’s 12 weeks against the same 12 weeks last year. Export queries with clicks, impressions, CTR, and position. Use the Search Console API or a connector if the 1,000-row interface limit cuts off your long tail.
- Drop branded queries into their own group. Brand terms behave differently. Amsive found only 4.79% of branded keywords triggered an overview, and those that did gained CTR.
- Tag intent. Informational (what, how, why, definitions), commercial (best, vs, alternatives, pricing, reviews), transactional or navigational (login, demo, the product name). A simple rule set on query words gets you most of the way.
- Tag overview presence using the method above.
- Hold position constant. Compare within position bands, such as 1 to 3, 4 to 10, and 11 to 20, so that a CTR change is not just a rank change.
- Read the grid. The pattern that points at AI Overviews is a CTR drop concentrated in informational queries with an overview, in the top bands, with flat positions and flat impressions. A drop spread evenly across every group points elsewhere.
This is the same logic Ahrefs used, a control group without overviews, applied to your own data. It shows correlation, not proof. You cannot rerun last year with the overview switched off.
What else causes rankings to hold while clicks fall?
Several things do, and each leaves a different trace. Rule them out before you report AI Overviews as the cause.
| Cause | What it looks like | How to check |
|---|---|---|
| Seasonality | Clicks and impressions move together across most query groups, matching last year’s shape | Compare the same weeks year over year, not this quarter against last |
| New or larger SERP features | CTR falls on queries with no overview, often where ads, a featured snippet, People Also Ask, video, or a local pack appeared | Check the SERP for your top losers; Amsive found queries with both an overview and a featured snippet fell 37.04% |
| Average-position artifacts | Average position looks steady while the mix of queries changed, or improves while clicks fall | Compare query by query and by position band, never on the site-wide average |
| Impression-counting changes | A step change in impressions and average position around mid-September 2025, when Google stopped honoring the num=100 results parameter | Do not compare periods that straddle September 2025 without noting it |
| Title link rewrites | CTR falls on a few pages whose title in the results no longer matches your title tag | Search the query and compare the shown title; Google documents when it rewrites title links |
| Competitor changes | One or two queries lose CTR after a rival adds rich results or ads | Look at the SERP for the specific queries |
| Tracking or consent changes | Analytics sessions fall but Search Console clicks do not | Compare Search Console clicks with analytics landing-page sessions |
Two of these need care. Average position in Search Console is the topmost position per query, averaged across queries, so a flat average can hide a page that moved from position 2 to 5 on its best query while gaining long-tail impressions elsewhere. And the num=100 change, reported widely by the SEO industry in September 2025 and never formally announced by Google, removed impressions that rank-tracking tools had been generating. Sites saw impressions fall and average position improve without any change in real visibility. A year-over-year comparison that crosses that date is not like for like.
Which pages lose the most clicks to AI Overviews?
Pages that answer a self-contained question lose the most, and pages a buyer has to visit to act lose the least. The studies support the informational half of that directly: Ahrefs’ and Seer’s samples are informational, and Pew found summaries on 60% of searches that began with a question word. The B2B mapping below is our reading of that evidence, not a measured result.
| B2B page type | Typical queries | Expected loss | Why |
|---|---|---|---|
| Glossary and definition pages | what is X, X meaning | High | The overview answers the question in a few sentences |
| How-to and explainer posts | how to do X, steps for X | High | Steps and lists summarize well |
| Statistics and benchmark roundups | X statistics, average X rate | High for the number, lower for the source | The figure gets lifted; a citation is the remaining value |
| Comparison and alternatives pages | X vs Y, best X for Y | Medium | Buyers often click through to check details |
| Pricing, product, and integration pages | X pricing, X integration with Y | Low | The answer depends on your current facts, and the buyer needs the page |
| Branded and navigational queries | your brand, your brand login | Low | Overviews appear rarely, and Amsive found branded CTR rose when they did |
| Demo, trial, and contact pages | book a demo, free trial | Low | The query is an action |
The table explains a common B2B pattern. Top-of-funnel blog traffic falls, pipeline from organic holds, and leadership sees the traffic chart first.
What should you do about clicks lost to AI Overviews?
Stop optimizing for clicks on queries where the click is gone, and spend the effort where it still exists or where exposure matters.
- Re-point effort to queries that still click. Rank the queries in your grid by clicks still available. Commercial comparison, pricing, integration, and use-case queries usually lead. Build and improve those pages first. This is ordinary SEO work, and it is where most B2B revenue from search comes from.
- Compete for the citation on the queries you keep losing. If an overview answers the question, the next best outcome is being one of its sources and being described correctly. AI Overviews cite pages Google retrieves for the query and for related sub-queries. How AI Overviews choose a source covers the method. Seer reported that cited brands had higher organic CTR than uncited ones, 0.70% against 0.52% in Q3 2025, and said it could not prove the citation caused it.
- Measure exposure separately from clicks. A citation that sends no visit is still read. Track AI feature impressions by page from Search Console and a logged query panel for what the answers say, as described in measuring AI search visibility. Score mention, citation, and accuracy apart from traffic.
- Change what leadership sees. Report clicks, AI exposure, and pipeline as three lines, and explain why the first can fall while the others hold. Attributing AI search to pipeline covers the analytics setup and the reporting format.
- Do not delete or noindex the losing pages. A definitional page that no longer earns clicks may still be the page an overview cites, and the page other pages link to. Removing it can cost citations and internal link value.
Do not try to keep the click by hiding the answer. Using nosnippet or max-snippet limits what Google can show in AI features and in normal snippets, which usually costs more than it saves.
What is the diagnostic checklist?
Work through it in order. Each step either rules a cause out or narrows the drop to a set of queries.
| Step | Question | Where to look | If yes |
|---|---|---|---|
| 1 | Did analytics sessions fall but Search Console clicks hold? | Search Console vs analytics landing pages | Tracking or consent issue, not search |
| 2 | Does the comparison cross mid-September 2025? | Your date ranges | Note the num=100 impression change before reading impressions or position |
| 3 | Does last year show the same seasonal dip? | Same weeks, year over year | Seasonality explains some or all of it |
| 4 | Is the drop concentrated in non-branded queries? | Branded vs non-branded split | Continue; branded drops point to demand or brand issues |
| 5 | Did positions hold within each position band? | Query-level positions by band | If not, it is a ranking problem first |
| 6 | Is the CTR drop concentrated in queries that show an overview? | Overview-present vs absent groups | AI Overviews are the likely cause |
| 7 | Did other SERP features appear on the losing queries? | Manual SERP check | Featured snippets, ads, or video share the blame |
| 8 | Do the shown titles match your title tags? | Manual SERP check | Title rewrites may explain specific pages |
| 9 | Are you cited in the overviews on your losing queries? | Manual check, logged | If not, that is the citation gap to work on |
| 10 | Do the losing pages still get AI feature impressions? | Generative AI performance report | Exposure held while clicks fell; report it |
Worked example: a diagnosis
This example is illustrative. The company and numbers are invented to show the method, and they are not a client result.
A B2B software company sees organic clicks down 22% year over year for September to November. Average position moved from 8.1 to 7.9. Impressions rose 3%. The marketing lead exports 1,800 non-branded queries for both periods, tags intent, and checks the top 150 by clicks lost for overview presence, signed out, on two days.
| Group | Share of last year’s clicks | CTR change, positions 1 to 3 | Clicks change |
|---|---|---|---|
| Informational, overview present | 41% | down 46% | down 44% |
| Informational, no overview | 12% | down 6% | down 4% |
| Commercial, overview present | 9% | down 14% | down 11% |
| Commercial, no overview | 23% | down 2% | up 5% |
| Branded (separate) | 15% | flat | up 2% |
The loss sits almost entirely in the first row: informational queries that show an overview, mostly glossary and how-to posts. The control rows are close to flat, so seasonality and a general demand drop do not explain it. The report periods do not cross September 2025, so the num=100 change is not a factor. Two commercial pages also lost CTR after a rival’s comparison page won a featured snippet, a separate problem with a separate fix.
The decision: shift the next quarter’s content plan from new glossary posts to integration and comparison pages, rewrite the top ten glossary pages so the answer leads each section and is easy to cite, and add AI feature impressions and cited-answer counts to the monthly report next to clicks.
Common mistakes
- Blaming AI Overviews from the site-wide chart. Fix: break the drop down by intent, overview presence, and position band before naming a cause.
- Comparing this quarter with last quarter. Fix: compare the same weeks year over year.
- Trusting average position. Fix: compare query by query within position bands.
- Comparing across September 2025 without a note. Fix: flag the
num=100impression change in any comparison that crosses it. - Quoting one study’s percentage as your forecast. Fix: use the studies for direction and your own grid for size.
- Counting a citation as a click. Fix: report AI exposure and clicks as separate numbers.
- Cutting pages that lost clicks. Fix: check whether they are cited or linked before you remove them.
A CTR drop with stable rankings is a measurement problem before it is a content problem. Once the grid shows where the clicks went, the plan follows: chase the clicks that remain, win the citations that replaced the rest, and report both. If you want the diagnosis done with you, that is part of an AEO engagement. The SEO, AEO, and GEO comparison explains how the three surfaces fit together.
Sources
- Update: AI Overviews reduce clicks by 58% — Ahrefs (Ryan Law), February 4, 2026, accessed 2026-11-08
- AIO impact on Google CTR: September 2025 update — Seer Interactive, November 4, 2025 (the page blocked automated access; figures checked against PPC Land’s report), accessed 2026-11-08
- Google AI Overviews reduce organic CTR 61%, paid traffic 68% — PPC Land (secondary, reporting Seer Interactive), November 5, 2025, accessed 2026-11-08
- Google AI Overviews study — Amsive (Will Guevara), April 16, 2025, accessed 2026-11-08
- Google users are less likely to click on links when an AI summary appears in the results — Pew Research Center, July 22, 2025, accessed 2026-11-08
- What are impressions, position, and clicks? — Google Search Console Help, accessed 2026-11-08
- Generative AI performance report (Search) — Google Search Console Help, accessed 2026-11-08
- Introducing Search generative AI performance reports in Search Console — Google Search Central Blog, June 3, 2026, accessed 2026-11-08
- AI features and your website — Google Search Central, accessed 2026-11-08
- Influencing title links in Google Search — Google Search Central, accessed 2026-11-08
- Google removes &num=100 parameter: what this means for your website — Locomotive Agency (Tyler Gargula), September 15, 2025, accessed 2026-11-08