TL;DR
- Two independent studies link AI Overviews to fewer clicks on Google’s traditional results, using very different methods.
- Steady rankings with falling clicks fit that pattern, but lower demand, a new query mix, other search features and tracking changes can look the same.
- Triage by query and page: check impressions, position, click-through rate and whether an AI feature appears.
- Keep useful informational pages. Change what they do for the buyers who still click, and judge decision-stage pages on qualified requests.
Why do clicks fall when rankings stay the same?
Often because the results page around your listing has changed. An AI Overview can answer part of a question above the links. Some searchers then stop, search again or leave. Your position holds, but fewer people need to click.
That explanation is plausible, and research supports it. For any single page, it is still a hypothesis until you check the queries behind the loss.
What do searchers do when an AI summary appears?
They click traditional results less often. Pew Research Center tracked the browsing of 900 US adults in March 2025 and checked which of their Google searches showed an AI summary. Pew Research Center, July 22, 2025
With a summary, people clicked a traditional result on 8% of visits, against 15% without one. They clicked a link inside the summary on 1% of visits.
That is an association, not proof of cause. Searches that trigger summaries may differ from those that do not.
Third bar: summary pages only. Pew Research Center, July 2025.
What does keyword-level data add?
It adds scale and a before-and-after view. Ahrefs compared 300,000 keywords in Google Search Console data: half with an AI Overview, half informational keywords without one. It tracked the top-ranked page’s desktop click-through rate (CTR) in December 2023 and December 2025. Ahrefs, February 4, 2026
Position-one CTR fell from 7.3% to 1.6% with AI Overviews and from 7.6% to 3.9% without. Ahrefs used the second group’s decline to model the first group’s expected CTR. Actual CTR was about 58% lower.
So the 58% is a model estimate for top-ranked desktop pages, not the raw drop or a share of all your clicks. CTR roughly halved without AI Overviews too, so not every decline is theirs.
How do the two studies compare?
They measure different things, so read them side by side rather than ranking them. Pew watches people. Ahrefs watches keywords. Both point the same way, which makes the pattern more credible than either study alone.
| Design point | Pew Research Center | Ahrefs |
|---|---|---|
| Unit measured | User visits to a results page | Top-ranked page’s average CTR per keyword |
| Sample | 900 US adults; 68,879 searches | 300,000 keywords in two equal groups |
| Period | March 2025; pages rebuilt in April | December 2023 vs. December 2025 |
| Comparison | Pages with vs. without a summary | Keywords with vs. without AI Overviews |
| Key result | 8% vs. 15% traditional-link clicks | CTR about 58% below a modeled baseline |
| Main limit | Observational; one month; Google only | Desktop, position one; model estimate |
Triage the loss before you change anything
Work query by query, not from a site-wide average. Export clicks, impressions, CTR and average position for the affected page’s queries, and compare the same weeks year on year. Then match the pattern to the table below.
Search Console’s generative AI performance report can help. It shows impressions of your links in Google’s AI features by page, country, date and device. Google says it reached all websites on August 31, 2026. Its listed views do not include queries, so pair it with your query export. Search Console Help
| What you see | Likely explanation | What to check | First response |
|---|---|---|---|
| Impressions fell with clicks | Less demand, seasonality or lost queries | Year-on-year impressions by query; changes in your market | Review demand and coverage first; impressions alone do not identify the cause |
| Average position fell | A ranking loss | Position by query, competing pages and recent site changes | Treat it as an SEO problem first |
| Impressions and position steady; CTR down on question-style queries | An AI Overview or another feature now answers part of the query | The live results page for those queries; AI-feature impressions for the page | Change the page’s job (see below) |
| CTR down on product or brand queries too | Ads, other result features or a changed title and snippet | The results layout, plus the title and snippet Google shows | Fix the snippet or review paid coverage before blaming AI |
| Search Console clicks steady, analytics sessions down | A tracking, consent or redirect change | Tag changes, consent rates and landing-page redirects | Fix measurement before touching content |
Worked example: a sizing guide that lost half its clicks
Worked example — fictional. An industrial filtration supplier sees clicks to its guide on choosing a filter micron rating fall by half, year on year. Its average position barely moved.
| Measure (same month, year on year) | Last year | This year |
|---|---|---|
| Impressions | 18,000 | 17,600 |
| Clicks | 1,260 | 610 |
| Click-through rate | 7.0% | 3.5% |
| Average position | 1.3 | 1.4 |
Demand and rank look stable, so the team splits the queries. Question-style searches, such as which micron rating suits hydraulic oil, carry most of the loss. A manual check shows an AI Overview on most of them. Searches that name a product series or ask for a supplier held steady. The generative AI report shows the guide earning impressions in AI features.
The diagnosis: the loss sits where AI Overviews appear. That makes them the most likely explanation, though not a proven cause. Nothing points to a ranking or tracking fault.
The team keeps the page. It adds what a summary cannot carry: a worked sizing example, the test standard behind its ratings and a link to request a compatibility check. It now reports the guide on requests started from the page. New writing time moves to the comparison and specification pages that buyers read before they ask for a quote. The team sets a review date and does not assume the clicks will return.
What should B2B teams change?
Keep useful informational pages. They still serve the buyers who click, and they can be sources for AI answers. Deleting them does not bring the lost clicks back.
Give the click a reason. Add what a short summary leaves out: worked examples, specifications, limits and a clear next step for a buyer with a specific case.
Report pages by their job. Judge informational pages on reach and the requests they start. Judge decision-stage pages, such as pricing, comparison and service pages, on qualified requests. The AI visibility measurement guide sets out counting rules.
Check what the answer says. If an AI Overview now answers for you, make sure it describes your offer correctly. The citation diagnostic guide shows how to trace a wrong statement to its source.
No change guarantees that clicks will return. Treat each change as a test, with a review date and a record of what changed.
AI Overviews and clicks: common questions
Do AI Overviews reduce clicks on every query?
No. The studies above describe averages, and many queries show no AI Overview at all. Check which of your own queries trigger one before you assume an effect.
Will a citation in an AI Overview make up for lost clicks?
Not reliably. A citation exposes a source link, which need not name your brand in the answer text, but links inside summaries drew few clicks in Pew’s data. Value a citation as visibility, and check its effect in your own data.
Does this evidence apply to AI Mode or chatbots?
Not directly. Both studies cover Google’s standard results pages. AI Mode and assistants such as ChatGPT have different layouts and send visits in different ways, so measure them separately.
Should we stop writing informational content?
Not by default. Write it when it answers a question your buyers ask and leads to a useful next step. Merge or remove pages that are thin or duplicated, not pages that are useful but now earn fewer clicks.
Turn the diagnosis into page changes
If your triage points to AI features on queries your buyers use, the next decision is which pages to change and how to tell whether it worked. SemShift’s generative engine optimization (GEO) service works from priorities agreed with your team. It revises existing pages and reviews cited sources alongside attributable visits and inquiries where those can be measured.
Bring the affected pages, a year-on-year query export and what your sales team hears from buyers who found you through search.