AI Overviews and Your Traffic: What the Evidence Shows

Yes, on the searches where they appear. Two randomized experiments and Pew's browsing data agree: fewer clicks reach websites, and almost none come from the links inside the summary. Google says total clicks are stable but has published no figures. Our stance: stop optimizing for the click you already lost on answer-style queries, and own the answer instead.

Do AI Overviews reduce clicks?

Yes. The strongest evidence is experimental, and the observational studies agree on direction, though not on size. The table ranks what we found from strongest to weakest evidence, and says who produced each number.

Read the numbers with care. Pew compared different searches, since Overviews favor longer, question-style queries, and re-ran the searches weeks later to see whether a summary appeared. Seer's 2025 update found that Overview queries already had lower click rates in June 2024, so part of the gap is query type, not the Overview.

Do AI Overviews send any traffic at all?

Yes, but little. In Pew's data, people clicked a link inside the summary in just 1% of visits to pages that showed one. Being cited still beats being skipped, though no public dataset tells you what one citation is worth.

• Cited versus uncited. Seer measured about 20,743 clicks per million informational impressions for cited brands and 9,445 for uncited ones. Seer cannot show that citation causes the gap, since stronger brands may simply get cited more.
• Early 2026 rebound. In Seer's cohort, organic click-through on Overview queries rose from 1.3% in December 2025 to 2.4% in February 2026. Seer calls it directional.
• Quality claim. Google says the remaining clicks are higher quality. The Agarwal and Sen experiment, as PPC Land reports, found that clicks returned by hiding Overviews looked equally engaged.

So the claim that lost clicks were junk is unproven. Check conversions by landing page before you decide a loss was harmless.

Which pages lose traffic and which hold up?

Question-style, comparison and quick-answer pages lose the most. Navigational, transactional and local queries are exposed far less, though not immune. Your site-wide loss depends on how many of your clicks come from searches that trigger an Overview.

• Most exposed. Pew found summaries on 60% of searches starting with who, what, when or why. Seer found Overviews on 95.4% of informational comparison queries and 85.9% of question queries.
• Less exposed. Seer saw Overviews on 36% of informational queries, 8% of commercial and 5% of transactional. Agarwal and Sen found a large effect on informational searches but none distinguishable from zero on navigational or transactional ones, from small samples.
• Do not trust "short is safe." Seer still saw Overviews on 27.3% of single-word informational queries.
• Publishers. Chartbeat data for 2,500-plus sites showed Google organic traffic down 33% worldwide and 38% in the U.S. between November 2024 and November 2025. The Reuters Institute says it is unclear how much is due to Overviews, with lifestyle and utility content more likely to be affected.

Our arithmetic shows why headline numbers mislead. If a third of your informational clicks come from searches with an Overview, and those lose 40%, informational clicks fall about 13%, not 40%. That is how Google can report stable totals while one site loses heavily.

What is zero-click marketing, and why does it matter now?

Zero-click marketing means earning attention and trust where people already get answers, without needing a visit. SparkToro, which says it coined the term, describes it as creating standalone value on platforms instead of dropping teasers and links, and not depending on referral clicks.

The pressure is measurable. SparkToro's 2026 study found 68.01% of U.S. Google searches ended with no click of any kind in January to April 2026, against 60.45% in 2024. The two years came from different panels, so treat the change as a direction. It also says traffic can fall while revenue rises, and still sees value in branded, local and transactional searches.

Our view: you cannot win back clicks that Google answers on the page. You can be the brand inside that answer. That covers the Overview, a ChatGPT or Perplexity answer that names you, and a helpful reply in a community your buyers read.

Publishers, who depend most on search clicks, are pushing back. In AEOeye's crawler access study (3 October 2026), 32 of 38 news publishers (84.2%) blocked at least one AI training bot, versus 12 of 193 SaaS brands (6.2%). It is a convenience sample, fetched once, and robots.txt is a request. Our reading: a brand is rarely hurt by blocking; it is hurt by being absent. Training-bot rules are a separate switch from appearing in Overviews, covered in what an AI Overview is.

How should you measure success when clicks fall?

Track impressions, citations and outcomes alongside clicks, and read clicks and impressions separately. A falling click-through rate can mean impressions grew, not that clicks fell. Google itself advises looking at conversions and engagement, not clicks alone.

The last row is what Search Console cannot show. AEOeye's free audit asks ChatGPT three unbranded buyer questions about your category and shows the full answers; the $29 report adds Google AI, Gemini, Claude and Perplexity. For daily monitoring at scale, dedicated trackers such as Profound, Peec AI and Otterly fit better.

What should you do this quarter?

Spend the quarter on five moves, in this order. They shift effort from the click you lost to the answer, the proof and the reason to visit.

1. Split your queries by intent. Group Search Console queries into question and comparison versus branded, transactional and local, then compare click change. Seer's advice applies: your own data beats aggregate benchmarks.
2. Triage your top pages by lost clicks. For each, add what a summary cannot flatten (original data, a tool, first-hand testing), merge duplicates, or accept the loss. We merged 165 of our own duplicate-intent pages into 42 for this reason.
3. Lead with the answer. Open each section with a self-contained 40 to 60 word answer that names your brand and the entity it covers. The AI Overviews playbook has the steps.
4. Give people a reason to click. Google says people still click in-depth reviews, original posts and first-person analysis. Add calculators, pricing, comparisons and checkers.
5. Report outcomes, not sessions. Put impressions, citations, branded search and conversions on the dashboard. For the wider picture, see AI search statistics and is SEO dead.

Clicks still matter on pricing, comparison and checkout pages. Everywhere else, the answer is the destination.