Do AI Overviews Reduce Clicks? What the Data Actually Shows (and How to Check Your Own Traffic)

Your rankings didn't move, your impressions look fine, and your traffic is down anyway. That pattern has a name now, and the question of whether ai overviews reduce clicks has stopped being a debate.

The answer is yes, and the size of it is no longer a guess. A randomized field experiment found that showing Google's AI Overviews cut outbound organic clicks by roughly 40% and pushed zero-click searches up by about a third. When researchers removed the overview for the same queries, outbound clicks jumped back up by 67.6%. Same users, same queries, same rankings. The only variable was the AI summary sitting on top of the results.

That is the cleanest evidence we have, because it is causal rather than correlational. Everything else in this article is about what it means for your specific traffic.

Yes, AI Overviews reduce clicks. Here is the short answer

Google AI Overviews reduce clicks to websites by roughly a third to a half on the queries they touch. The strongest evidence: Pew Research found that users clicked a traditional result in only 8% of visits when an AI summary appeared, versus 15% when no summary showed. That is a relative decline of about 47%.

The mechanism is not mysterious. The AI summary answers the question directly on the results page, so the user never needs to leave Google. The click that used to go to your page gets consumed as an impression on the SERP instead.

This is why your rankings can hold while your traffic falls. Position one still means position one. It just means less than it used to when an AI Overview sits above it, because the overview absorbs the intent before the user ever reaches the blue links.

The confusion comes from expecting a rankings problem. This is not a rankings problem. It is a click-through problem, and those need different diagnostics and different fixes.

What the studies actually measured (and where they disagree)

Three evidence sources matter here, and they measured different things, which is why the numbers look inconsistent until you line them up.

Pew Research used behavioral panel data, meaning they watched what real users actually did in their browsers. Their finding, the 8% versus 15% click rate, is the cleanest measure of user behavior change. It tells you what people do when an AI summary appears, not what a ranking tool estimates.

Ahrefs ran a different analysis. They compared click-through rates across 300,000 keywords and found AI Overviews correlated with a 34.5% lower average CTR for the top-ranking page on informational keywords. Later, comparing December 2023 with December 2025, they reported the gap had widened to 58% lower CTR for top-ranking pages. Same methodology, two years apart, and the effect grew.

The randomized field experiment is the strongest of the three because it controlled for everything except the overview itself. When the overview was removed, zero-click searches fell from 0.73 to 0.54 per search. That tells you the mechanism is not "users got less curious." It is that the answer was already on the page, so there was no reason to leave.

Where the studies disagree is mostly about magnitude and trend. The Ahrefs data suggests the effect worsened through 2025. Seer Interactive's time series showed CTR on AI Overview results bottoming at 1.3% in December 2025 and rebounding to 2.4% by February 2026, which hints at partial stabilization rather than a one-way collapse.

Treat any single number as a snapshot, not a permanent state. The direction is settled. The exact size keeps moving.

Which queries get cannibalised, and which are safe (for now)

Not all queries lose clicks equally. The damage concentrates in informational, question-shaped searches where the full answer fits in a paragraph: definitions, how-tos, simple comparisons, factual lookups. These are the queries where Google's AI can generate a complete, confident answer without needing the user to visit anyone's site.

Transactional queries are safer. Someone searching "buy running shoes size 11" or "dentist near me open Saturday" still needs to click through to a product page or a booking form. The AI Overview can summarize options, but it cannot complete the transaction.

Navigational queries are safer too. If someone types your brand name, they are looking for your site specifically, and no AI summary will satisfy that intent.

The practical move is to classify your own keyword list. Pull your top 50 queries from Google Search Console. Mark each one as informational, transactional, navigational, or local. The informational ones are where your CTR decline will show up first, and where the traffic drop will be steepest.

This is also where the query mix matters more than any single study. A site that gets 80% of its traffic from informational content will feel the AI Overview cannibalisation far harder than a site that mostly ranks for product and brand terms. The aggregate studies average across everything. Your exposure depends on your mix.

The citation exception: being the source AI quotes changes the math

Pages cited inside an AI Overview can earn comparable or even higher CTR than a standard blue link, because the citation carries an endorsement. The traffic loss falls hardest on pages that rank but are not cited.

This is the nuance that most panic pieces skip. When Google generates an AI Overview, it pulls from specific pages and links to them as sources. Those citation links still get clicked. They get clicked by users who want to verify, go deeper, or check the original context. And because the AI has already vouched for the source, the click that does happen tends to be higher intent.

The reframe matters. The problem is not "AI Overviews are stealing traffic." The problem is that uncited rankings are being devalued. A page-one position that used to guarantee a share of clicks now guarantees much less, unless you are one of the two or three sources the AI chose to quote.

That means the fix is not to abandon informational content. It is to restructure it so the AI wants to cite you. Clear, direct answers near the top of the page. Specific data points. Original framing that the AI cannot generate on its own. The full playbook for that is in our guide on how to optimize for Google AI Overviews and get cited.

How to check your own traffic drop in Google Search Console

Open Google Search Console, go to Performance, and set the date comparison to the last 16 months. You are looking for a specific signature: impressions stable or rising, position stable, but CTR falling on your informational queries. That pattern is the fingerprint of AI Overview cannibalisation.

Filter by query type. Click into your top informational queries, the ones that are question-shaped or definition-seeking. Compare their CTR over the past six months against the same period a year ago. If impressions held but CTR dropped 30% or more, and your position did not change, the AI Overview is the most likely cause.

Rule out other causes before you blame the overview. Google's own documentation on diagnosing search traffic drops walks through the technical checks: manual actions, security issues, crawl errors, algorithm updates. Run through those first. If none of them explain the drop and the CTR pattern matches, you are looking at AI Overview impact.

Segment by page, not just by query. Some pages will be hit harder than others depending on whether their target queries trigger an overview. A page that ranks for five informational queries might lose 40% of its clicks while a similar page ranking for transactional terms loses nothing. The aggregate number hides this. The page-level view shows it.

For a deeper workflow on tracking this over time, we have a separate guide on using Google Search Console to track AI-driven traffic.

Why Search Console won't show you AI Overview data directly

Google folds AI Overview impressions and clicks into the same "Web" search type in Search Console, with no filter to separate them. You cannot pull a report that says "here are the clicks you lost to AI Overviews." You can only infer the impact indirectly, through the CTR-delta method described above.

This is a deliberate product decision, and it frustrates everyone who tries to measure the problem precisely. Google knows exactly which queries trigger an overview and which clicks came from citation links versus blue links. They just do not share that breakdown.

Some third-party tools attempt to estimate AI Overview exposure by scraping the SERP and checking which queries trigger an overview. That gives you a directional read, but it is still an estimate. The only ground truth Google provides is the aggregate Performance report, and it lumps everything together.

Set your expectations accordingly. You can measure the effect. You cannot measure it with the precision you might want. The CTR-delta method will tell you whether the problem is real and roughly how big it is. It will not tell you exactly which clicks went to the AI summary and which went to a cited source.

What to actually do about it (without abandoning SEO)

The worst response to AI Overviews reducing clicks is to stop creating informational content. Informational queries still drive discovery, brand awareness, and the kind of topical authority that makes Google trust your site in the first place. The better response is to change what you expect from that content and how you structure it.

Start by identifying which of your affected pages can realistically win the citation slot. Look at the pages that lost CTR but still rank in the top five. Those are your candidates. The AI is already pulling from pages in that range. The question is whether it pulls from yours or a competitor's.

Restructure those pages so the answer is quotable. Put the direct answer in the first two sentences. Use specific numbers, named frameworks, and original data where you can. The AI prefers sources it can quote cleanly, with minimal rewriting. A page that buries the answer in paragraph seven gives the AI nothing to work with.

Shift some effort toward query types the overview does not absorb well. Comparison pages with real testing, opinion pieces with a clear point of view, tools and calculators, local content with specific details. These still earn clicks because the AI summary cannot replicate them.

Treat the lost clicks as top-funnel impressions. Someone who reads your brand name inside an AI Overview has still encountered you, even if they did not click. That exposure has value, but you need to measure it differently. Track branded search volume over time. If your informational content is getting cited, you should see branded queries rise even as raw CTR falls.

If you have older posts that lost traffic, a refresh workflow that recovers lost traffic can bring some of it back, especially if you restructure the content for citability at the same time.

The bigger shift: clicks were a proxy, and the proxy is breaking

AI Overviews are the visible edge of a broader move from click-based discovery to answer-based discovery. ChatGPT, Perplexity, and Gemini are doing the same thing Google's overview does: answering the question directly and deciding which sources to name. The metric that replaces raw CTR is whether you are the recommendation, and that is measurable.

This is where the mental model needs to update. For twenty years, the click was the unit of search value. You ranked, you got clicked, you got a chance to convert. That chain is breaking at the middle link. The ranking still matters. The conversion still matters. The click is becoming optional.

The sites that adapt fastest are the ones that stop measuring success by organic sessions alone and start measuring whether AI systems name them when buyers ask. That is a different metric, and it requires different tracking. Google Search Console will not show it to you, because Google has no incentive to make the cannibalisation easy to see.

Running these measurements against live AI engines at AnswerRank, the pattern that keeps surprising us is how often a brand ranks well on Google but is completely absent from the AI answers for the same queries. The overlap between "ranks on page one" and "gets cited by AI" is smaller than most site owners assume. That gap is where the traffic went.

The free GEO Rank Tracker shows whether ChatGPT, Perplexity, and Gemini name your brand when someone asks the questions your buyers ask. It takes two minutes and gives you a baseline you cannot get from Search Console.

The question of whether ai overviews reduce clicks has a clear answer: they do, by a third to a half on affected queries, and the effect is growing. The more useful question is whether your pages are the ones still getting cited and clicked. Check your CTR deltas in Search Console this week, restructure your top informational pages for citability, and start tracking whether AI names you at all.