Google's AI Overviews: what the browsing data really shows
AI Overviews, the AI-generated summaries at the top of Google's results, reduce clicks to websites. The Pew Research Center, which tracked the real browsing of 900 US adults, measures an 8% click rate when a summary is present versus 15% when it is not. The loss is real, but it falls mostly on traffic that was never going to buy, and the wrong reflex costs more than the decline itself.
- The click rate roughly halves when an AI Overview appears, measured on real browsing rather than on what people report.
- Being cited in the overview brings almost no traffic. The value sits elsewhere.
- Commercial queries are far less affected than informational ones.
- Not appearing among the sources means disappearing twice, from the classic result and from the generated answer.
- Before concluding you have lost revenue, check whether inquiries fell as much as sessions did. In most cases, they did not.
On this page
AI Overview
An AI Overview, Google's name for an AI-generated summary, is a block of text produced by a language model and displayed above the classic search results. It synthesizes several web sources and cites them with links. It answers the query directly rather than pointing to a page, which changes the relationship between the engine and the sites it indexes.
What the research actually measures
For an executive team, the question is not how many clicks Google absorbs. It is whether inquiries and revenue have fallen, and whether to reallocate a budget. Both answers are often reassuring, provided you look at the right numbers, because the decline falls overwhelmingly on traffic that was never going to buy or ask for anything.
If you have seen your sessions drop and your first reflex was to publish more articles, that is the move that makes things worse. A drop in traffic reads like a drop in store footfall: what matters is not the number of people who walked in, it is the number who bought, and the two do not move together.
The topic has produced many claims and few measurements. The Pew Research Center study is the most solid one available, because it observes real behavior rather than asking about it.
The center tracked the browsing of 900 US adults who agreed to install a tracker, over the month of March 2025. The dataset covers 68,879 distinct Google queries, of which 12,593 triggered an AI Overview.
| Measure | With an overview | Without one |
|---|---|---|
| Click on a classic result | 8% of visits | 15% of visits |
| Click on a cited source | 1% of visits | Not applicable |
| Session ends after the page | 26% | 16% |
The third row is the one quoted least and the one that weighs most. A user who meets an AI Overview ends their browsing session far more often. This is not a customer lost to a competitor, it is a person who got their answer and was buying nothing. The overview does not redirect traffic elsewhere, it absorbs it. For a business, the consequence is clear: part of the lost traffic goes nowhere, and no advertising budget will buy it back.
Pew states its own limits: the panel covers one particular slice of US adults, the measurement spans a single month, and Google's product changes constantly. Treat the orders of magnitude as a direction, not as constants. No budget decision should rest on the 8% figure alone, but on your own curves.
What to check before you react
A drop in sessions is not a drop in revenue. Commercial queries and vendor searches are barely affected, because those decisions require visiting sites. What disappears is mostly informational traffic, the kind that almost never produced an inquiry or a sale. Before committing anything, you need to know which of the two curves moved.
- Did our inquiries and revenue fall in the same proportion as our sessions, over the same period?
- Does the decline hit our service pages or only our articles?
- Which queries actually brought us customers, and are they affected?
- Are we cited as a source in the generated answers on our topics, and who checks?
- What would we do differently if this traffic never came back?
The useful answer separates the two curves and names the affected pages. A vague answer talks about the end of SEO, or proposes increasing publishing volume to compensate.
Deciding whether to reallocate a budget or change nothing is a trade-off worth tens of thousands of dollars a year. A 90-minute consultation settles it on your real data, with a written summary that circulates in your organization.
Anatomy of an AI Overview
Pew's data also describes the shape of these blocks, which helps you understand how to appear in them.
| Characteristic | Measure |
|---|---|
| Median length | 67 words |
| Shortest observed | 7 words |
| Longest observed | 369 words |
| Overviews citing three or more sources | 88% |
| Overviews citing a single source | 1% |
| Links to government sites | 6%, versus 2% in classic results |
Two operational lessons follow. First, the median length of 67 words explains why short, self-contained answer blocks get picked up: they match the format the AI assembles. Second, 88% of overviews cite at least three sources, which means the spot to occupy is not a single one. The cost of entry is therefore far lower than that of a classic first position, and that is the strongest argument for investing now rather than waiting.
Pew notes finally that Wikipedia, YouTube and Reddit dominate the citations, in overviews as in classic results.
What is affected, and what is far less so
Treating the decline as uniform leads to cutting or investing in the wrong place. The effect varies sharply with what the person is looking for.
| Query type | Example | Exposure to the overview |
|---|---|---|
| Definition and explanation | What is internal linking | Very high, the click often disappears |
| Single fact | GST rate in Quebec | Very high |
| Comparison | WooCommerce or Shopify | Moderate, the reader wants detail |
| Vendor evaluation | SEO agency Montreal | Low, the decision requires a visit |
| Transaction | Website design quote | Low |
| Brand navigation | Company name | Near zero |
A business whose blog loses informational traffic while its service pages hold steady is not facing a crisis. It is facing a shift. The reflex of compensating by publishing more informational articles makes the problem worse rather than fixing it, and it commits a budget to exactly the kind of content the overviews absorb best.
What stays in-house is the list of queries that bring in customers, not the ones that bring in traffic. No one on the outside knows which inquiries turned into engagements or which were worth it, and that list is what decides whether the decline is serious or not. What gets delegated is measuring your presence in the generated answers: tracking citations, share of voice by topic and the quality of referred traffic takes tools and a weekly read that an internal team will not sustain over time.
How to respond without fighting the wrong battle
Three misreadings are going around, and each one translates into a budget committed in the wrong place.
First mistake: concluding that SEO is dead. Overviews draw from web pages. Not appearing among the sources means disappearing from both the classic result and the generated answer. Organic visibility now feeds two channels instead of one, which increases the return on the same budget rather than reducing it.
Second mistake: aiming for the citation for the traffic it brings. It brings almost none, 1% of visits according to Pew. Its value is elsewhere: being named as a source in an answer builds a recognition that shows up later, on a brand query or a direct contact. Budgeting the citation as a traffic source leads to guaranteed disappointment.
Third mistake: measuring success in clicks alone. If the goal becomes getting cited, the indicator becomes presence in the overviews and in assistant answers, not just referral traffic.
Shift the effort toward commercial queries
They are barely affected and they are the ones that bring engagements. An informational article that loses 40% of its traffic deserves less investment than a service page that does not yet exist. It is the only budget shift the situation truly justifies.
Structure for extraction
A direct answer of 40 to 60 words at the top of the page, self-contained sections, sharp definitions. It is the format the AIs assemble, and it also serves the human reader.
Produce what an AI cannot fabricate
Clean data, real cases, a clear-cut position. Generic content feeds the overview without ever being cited by name. Content that brings a unique data point becomes the source.
Track citations, not just clicks
Presence in the generated answers can be measured. It becomes an indicator in its own right, alongside rankings and traffic. To date, no one in the industry can honestly link a citation to a sale, and a vendor who promises it is selling a certainty that does not exist.
The AI Overview does not replace the web. It reads it. Not being readable by it means disappearing twice.
Falia analysis gridThe six points that follow can be checked in an hour and determine whether there is a decision to make or not.
Diagnosing your exposure
The method for structuring content to be citable is detailed in structuring content to be extracted. The economic effect on the value of content is covered in the collapse of production cost. The general framework of SEO is in the four layers of search. The execution is on the GEO and AEO page.
Before blaming the summaries, the five documented causes of a traffic drop are reviewed in the five documented causes of a site with no traffic.
The overall view of AI visibility is in who measures what, and who measures nothing.
Visibility in answer engines is at the heart of the Strengthen your visibility in AI answers goal.
Already running a marketing team? See how we plug in as reinforcement on answer engine optimization.
Frequently asked questions about AI Overviews
Do AI Overviews really reduce traffic?
Yes, and it is measured on real browsing. The Pew Research Center tracked 900 US adults in March 2025 and observes an 8% click rate to a classic result when an AI Overview is present, versus 15% when it is not.
Does being cited in an AI Overview bring traffic?
Very little. Only 1% of visits lead to a click on a source cited in the overview, according to the Pew Research Center. The value of the citation is the recognition earned with a customer in research mode, not the direct traffic it generates.
What share of searches triggers an AI Overview?
About 18% of Google searches in March 2025, according to the browsing data analyzed by the Pew Research Center. Six users out of ten encountered at least one AI Overview during the month observed.
Are all queries affected the same way?
No. Definition and single-fact queries are highly exposed, because the AI Overview answers them completely. Vendor evaluation queries and transactional queries are far less so, since the purchase decision requires visiting sites. That distinction is what decides whether to react or not.
Has SEO become useless?
No, the opposite, and that is what should guide the budget decision. AI Overviews are built from indexed web pages. Not appearing among the sources means disappearing from the classic result and the generated answer. Organic visibility now feeds two channels instead of one, for the same budget committed.
How do you improve your chances of being cited?
By structuring content for extraction: a direct answer of 40 to 60 words at the top, self-contained sections, sharp definitions and sourced figures. And by publishing what an AI cannot fabricate, namely clean data and real cases.
- Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, browsing study of 900 US adults, March 2025, published July 22, 2025.
- Pew Research Center, Do people click on links in Google AI summaries?, accessed July 2026.
- Google Search Central, AI features in Google Search, accessed July 2026.
Geneviève puts the strategy for your engagement into action. She leads all our web development projects: Shopify, WordPress and the new ways of building a site with AI. She manages our team of developers and translates your business needs into technical language. She runs your organic search (SEO), your visibility in AI answers (GEO) and your site's conversion rate optimization (CRO). Her work is at the heart of three goals: Attract customers with SEO and AI, Improve your site's conversion, and Strengthen your visibility in AI answers. With Gabriel, she also builds the landing pages for your advertising campaigns. She writes mainly about SEO, AI visibility and web design.
About Falia →
SEO
Paid advertising
Conversion
AI visibility
Web design