Google's AI Overviews are answering your readers' questions before they reach your site
A compliance guide you published eighteen months ago used to be a reliable source of new visitors: it ranked on the first page for a handful of long, specific questions, and a steady trickle of people landed on it every week. The ranking has not moved. Search Console still shows the page in position two or three for the same terms. What changed is that the number of people who actually click through has quietened to almost nothing, and nothing in the dashboard explains why, because nothing about the page itself is wrong.
What happened sits above the page, not inside it. Google now answers a large share of informational searches directly on the results page itself, through a feature called AI Overviews: a short, AI-written summary that sits above the list of blue links, built from several sources at once. The visitor reads the summary, gets enough of an answer to stop looking, and never scrolls down to the link that would have brought them to the site that did the actual work of researching and writing the explanation.
This is not a glitch or a temporary dip to wait out. It is a structural change in how a large share of search traffic behaves, and the kind of informational, question-shaped content a classifieds operator relies on for organic reach sits squarely in the part of search that this change affects most. Understanding exactly what the data shows, rather than reacting to a vague sense that "search traffic is down," is what decides whether the next year of content spending gets allocated somewhere that still works.
What Google's own usage data actually shows
The most carefully controlled study of this came from the Pew Research Center, which tracked the real browsing activity of 900 U.S. adults in March 2025, covering 68,879 individual Google searches, rather than relying on surveys or estimates. An AI Overview appeared on 18% of those searches. When one did, the visitor clicked through to a normal search result in just 8% of visits. When no AI Overview appeared, that click rate nearly doubled, to 15%. Clicking on a link inside the AI Overview itself, the sources it was built from, happened in only 1% of visits where a summary showed up.
The same research found that a visitor was more likely to simply stop searching altogether after seeing an AI Overview: 26% of visits to a page with a summary ended the browsing session outright, against 16% on pages without one. Put together, the picture is not that AI Overviews send traffic to different places. It is that a meaningful share of searches that used to end in a click now end in nothing, because the summary already answered the question well enough. Google disputed Pew's methodology on the day the study came out, calling the sample unrepresentative of search traffic as a whole and maintaining that it had not seen a significant drop in aggregate web traffic; that counterargument is worth knowing, but it does not change what the browsing data shows for the specific pages and query types the study actually tracked.
Independent analysis outside Pew's study points the same direction with larger, if less tightly controlled, numbers, and the trend across them is toward a bigger effect, not a smaller one. Ahrefs ran the same comparison twice: in April 2025, using a year-over-year sample of 300,000 keywords, it found the position-one organic result losing roughly 34.5% of its clicks on searches where an AI Overview appeared; re-run on December 2025 data, the loss had grown to about 58%. DMG Media, the publisher behind the Daily Mail and Metro, told the UK's Competition and Markets Authority that across a six-week sample of its own top-ranked pages, desktop click-through fell from 25.23% on searches with no AI Overview to 2.79% on searches where one appeared above a visible link to its site, a drop of 89%, with mobile down by a comparable 87%. Chegg, an education platform whose business depends on people searching for answers to study questions, reported a 49% decline in non-subscriber traffic over a single year and has since sued Google over it.
None of these figures describe a platform in decline. Google's own search volume is still growing. What they describe is a shift in where the click goes once the search is answered, and for a site whose content exists specifically to answer a question a reader typed into Google, that shift lands directly on the traffic the content was built to earn.
Why explainer content is exactly what gets summarized
The same Pew data explains, in detail, which kinds of searches are the ones most likely to trigger a summary in the first place, and the pattern is not flattering for a content strategy built around long-form guides. A search of one or two words produced an AI Overview only 8% of the time. A search of ten words or more produced one 53% of the time. A search phrased as a question, starting with "who," "what," "when" or "why," triggered a summary 60% of the time. A search written as a full sentence, with a subject and a verb, triggered one 36% of the time.
That is a precise description of the queries a compliance or operations blog is written to catch: "what happens when a payment processor freezes a merchant account," "how long does an age verification record have to be kept," "what counts as a doorway page under Google's spam policy." These are exactly the long, question-shaped searches Pew found most likely to produce a summary, which means the content format that has worked well for organic reach over the past few years is also the format Google's own system is now built to intercept before the click happens.
This does not mean writing the content was a wasted effort. The median AI Overview in Pew's sample was 67 words long and cited three or more sources in 88% of cases, which means the underlying research, the actual explanation of how a payment network or a verification law works, still has to come from somewhere, and a well-sourced, carefully reasoned article is a more likely candidate to be cited than a thin one. The cost is not that the work stopped mattering. It is that being cited and being clicked have become two different outcomes, and only one of them still shows up as a visitor on the site.
It is also worth being precise about what this does not cover. The data behind this shift comes from informational, question-based search, the kind that asks how something works or what a law requires. A search with clear commercial or local intent, the kind a visitor types when they already know what they want and are deciding where to get it, behaves differently, and nothing in this research suggests that category of search is being hollowed out the same way. The risk is specific to the explainer layer of a site's content, not to the site's core business.
The one kind of search that still converts into a click
There is a clear exception in the data, and it points toward the one defense that actually holds up over time. Research from the marketing agency Amsive found that branded searches, meaning a search that already includes a company or product name, saw click-through rates increase by around 18% when an AI Overview appeared, even as the generic, non-branded searches described earlier kept losing a third or more of their clicks over roughly the same stretch.
The mechanism is straightforward once it is stated plainly: an AI Overview that names a specific brand is functioning as a form of endorsement, telling the reader that this particular source is the relevant one, and a reader who already has some familiarity with that name is more inclined to go check the source directly rather than take the summary's word for it. A generic question has no such anchor, so the summary simply absorbs the answer and the reader moves on without ever learning, or caring, which site supplied it.
This is also why chasing a spot inside the AI Overview itself is a weaker strategy than it sounds. Research from Authoritas tracking which pages get cited inside AI Overviews over time found that roughly 70% of cited pages changed within a two-to-three month window, with little relationship to how those same pages were doing in ordinary organic rankings. A citation today is not a stable asset; it is closer to a lottery result that gets redrawn every few weeks, for reasons that do not track the normal SEO signals a site can actually control.
What a site can control is whether its own name is recognizable enough that people start typing it directly, rather than typing the generic question and discovering the site by chance through a ranking that Google's own summary increasingly stands between. That is a slower, less exciting kind of SEO work than publishing another explainer, but it is the one category of click this data shows is actually growing rather than shrinking.
What this changes about where content budget should go
The practical shift is not to stop publishing explainer content, since a classifieds operator working without access to Google or Meta advertising still needs organic search to carry a real share of new visitor discovery, and abandoning that channel is not a realistic option. The shift is in what gets measured and what gets expected from it. A page that still ranks but no longer converts into clicks is not a failed page; it is a page doing a different job now, informing the AI Overview's answer rather than pulling in a reader directly, and treating its traffic number as a failure misdiagnoses what is actually happening.
Judge content going forward by the gap between impressions and clicks in Search Console, not by raw position. A page holding its ranking while its click-through rate collapses is showing the exact symptom this shift produces, and that symptom calls for a different response than a page that genuinely fell out of favor with the ranking algorithm. Rewriting the second kind of page with better keywords helps. Rewriting the first kind rarely does, because the problem sits above the page, not inside it.
Put the content budget that would have gone into a fourth or fifth version of an already-covered explainer topic into channels where the click still reliably happens: an email list built from visitors who already converted once, a community or forum presence where a real person answers a real question instead of an algorithm summarizing one, and the brand-recognition work that Amsive's data shows is the one search behavior still moving in the right direction. A directory's city and category pages belong in this plan too, but as commercial pages built around real, current listings, not as a second front of informational content competing for the same question-shaped searches that AI Overviews already dominate.
Keep writing explainer content where a genuine gap exists, since being one of the three or more sources an AI Overview cites is still worth something: it is a credibility signal, a link some readers do follow, and raw material for the brand recognition that converts into direct searches later. Just stop budgeting for it as if the click were guaranteed, because the data from the past two years says plainly that it no longer is, and a content plan built on that old assumption will keep producing pages that rank perfectly and get read by almost no one.
