Public research agrees on direction and disagrees on size. Pew Research Center found people clicked a standard result in 8% of visits that showed an AI summary and 15% of visits that did not. Ahrefs and Seer Interactive found larger drops for specific query sets. Google says overall click volume is stable. Each number answers a different question.
Pew: what real browsing looked like
Pew Research Center used the browsing data of 900 U.S. adults from March 2025. It reran their Google searches in April and recorded which results had an AI summary. Of 68,879 unique searches, 12,593 produced one.
Two more figures from the same study. Visits with an AI summary ended the browsing session 26% of the time, against 16% without one. And around two-thirds of all searches, with or without a summary, ended without a click on a result.
The limits matter. The summaries were collected weeks after the searches, and Pew notes a summary can change over time. It counted only the first 3 cited links, and when a link appeared in both the summary and the results it credited the summary. Most important, the two groups are not matched. Pew found that longer searches and searches phrased as questions produced summaries far more often, with 60% of searches starting with a word like “who,” “what,” “when,” or “why” getting one. People asking those questions may have clicked less anyway.
Ahrefs: the top result, modeled
Ahrefs compared click-through rates for the number 1 result across 300,000 keywords: 150,000 where an AI Overview appeared and 150,000 informational keywords where one did not. It used aggregated Search Console data for desktop. Its re-run, published Feb 4, 2026, compared December 2023 with December 2025.
Here is the arithmetic. For informational keywords with no Overview, the top-result click-through rate fell from 0.076 to 0.039. For keywords that had an Overview by December 2025, it fell from 0.073 to 0.016. Ahrefs applied the first drop to the second group to estimate what would have happened without an Overview, got 0.037, and called the gap from 0.016 a 58% reduction.
That 58% is a modeled difference, not a measured one. It assumes the two groups would have declined at the same rate. It covers desktop and the top position only. It also differs sharply from the 34.5% Ahrefs reported from the same method in April 2025, which shows how much one assumption and one more year can move the answer. Ahrefs also sells AI visibility tools, which is worth knowing when you weigh the framing.
Seer: citations and click-through
Seer Interactive looked at 3,119 informational queries across 42 client organizations, from June 2024 to September 2025, covering 25.1 million organic impressions. For the third quarter of 2025 it reported organic click-through rates of:
- 0.52% where an AI Overview appeared and the brand was not cited.
- 0.70% where an Overview appeared and the brand was cited.
- 1.45% where no Overview appeared.
Seer’s own caveat is the useful part. It says it cannot prove citation causes the higher click-through rate, because brands with stronger authority may be both cited more and clicked more. It also found that queries that later triggered an Overview already had lower click-through rates than other queries in every month it tracked, and that queries with no Overview declined too. Something broader than the Overview itself is moving clicks.
The sample is client websites, not a random cross-section of the web, and it covers informational queries only.
Google’s position
In an Aug 6, 2025 post, Google’s head of Search said total organic click volume from Google Search to websites has been “relatively stable year-over-year” and that average click quality has increased. It described third-party reports of dramatic declines as often based on flawed methodologies or isolated examples. The post gave no figures, so there is nothing to check it against. It also said traffic is shifting between sites, which fits both a stable total and big losses for some publishers.
How to check your own pages
You can get a usable answer from Search Console in an afternoon. Google says sites that appear in AI features are included in the overall search traffic reported there, under the Web search type, and its guide for AI features also points to a Generative AI performance report. Check what that report covers for your property before you rely on it.
- Pick 20 or so informational queries where you rank in the top 5 and that bring real traffic.
- Search each one and note whether an AI summary appears and whether your page is cited in it.
- For the queries with a summary, compare clicks and click-through rate for equal periods before and after it started showing. Use the same weeks a year earlier to take out seasonality.
- Run the same comparison on similar queries with no summary. That is your control group.
- If the two groups moved by about the same amount, the Overview is probably not your problem. If only the first group fell, you have your answer for those pages.
This is still a correlation. It is just the correlation that matches your site, your queries, and your audience, which is more useful than a headline average.
How to read all of this
The sources are not contradicting each other as much as they look. Pew compares visits across all searches. Ahrefs isolates the top result on informational keywords. Seer follows a fixed set of client queries. Google speaks about the total across all sites. A site that publishes mostly short-answer informational content can lose a lot while the aggregate barely moves. Seer limited its study to informational queries because it considers them the most exposed, so its numbers are not a guide for other query types.
For your own site, the number that matters is yours. Look at clicks and impressions in Search Console for the pages and queries where an Overview shows, before and after, and compare with similar queries that have none. That gives a real answer for your pages, with the same caveats about causes.
What the studies cannot tell you is whether a click you did not get turned into a visit later, or whether a brand mention in the summary did some of the work a click would have done. That is a reason to track whether you are cited and mentioned in the answer, not only whether anyone clicked. Our write-up of how AI Overviews work covers the tracking side, and the pattern in AI Overview clicks fall describes how to confirm it on your own data.
