Zero-Click in the AI Era: What Pew, Ahrefs and Seer Actually Measured

What did each study actually measure?

Zero-click searches aren't new — Google's own featured snippets and knowledge panels have answered questions directly in results for years. What's new since AI Overviews rolled out broadly is the scale, and three studies get cited constantly as evidence of impact. They don't measure the same thing.

Pew Research Center (22 July 2025, based on March 2025 browsing data from 900 US adults and 68,879 Google searches) measured click rate per visit: a traditional result link was clicked in 8% of visits where an AI summary appeared, versus 15% of visits where it didn't. About 1% of visits ended in a click on a link inside the summary itself. Visits with an AI summary also ended the session (no further click at all) 26% of the time, versus 16% without one. This is behavioral data from real people's browsing, not a simulated query set.

Ahrefs (11 August 2025, Louise Linehan, 15,000 long-tail queries via Ahrefs Brand Radar) didn't measure clicks at all — it measured whether a URL cited by an AI assistant also ranks in Google's top 10 for the same query. Only about 12% do, on average, with Perplexity a clear outlier at 28.6% and ChatGPT, Gemini and Copilot all around 6–8%. This is a ranking-overlap study, often misquoted as a clicks study because it sits next to clicks data in roundups.

Seer Interactive (Tracy McDonald, published 4 November 2025, covering 42 client organizations, 3,119 search terms and 25.1 million organic impressions from June 2024 through September 2025) measured CTR trend over time specifically on queries that triggered an AI Overview. Organic CTR on those queries fell from a mid-2024 baseline to a floor of 0.61% by around September 2025, while non-AI-Overview queries also declined, though less steeply, to about 1.62% in the same window. Seer's team is explicit that they "cannot definitively prove that citation causes higher CTR" for the brands they studied — stronger brands may simply get cited more often in the first place.

Why do the numbers differ?

Different samples, different countries, different time windows, and — most importantly — different metrics entirely. Pew's 8%/15% split is a snapshot from one country in one month. Seer's 0.61%/1.62% figures track a trend across 19 months on a specific client roster, mostly informational and educational queries, which may not generalize to commercial or transactional searches. Ahrefs isn't measuring clicks in the first place, so stacking its 12% next to Pew's 8% as if they're comparable "drop" figures is a category error that shows up often in roundup posts. Treat any single blended number — "clicks are down 30 to 40%" — with suspicion: it doesn't trace to one publisher, sample and date. The real, sourced figures in this space are actually higher and narrower than that range. A pre-registered randomized field experiment by researchers at the Indian School of Business and Carnegie Mellon University (1,065 US desktop Chrome users, Jan–Feb 2026, reported by Search Engine Journal on 27 Apr 2026) found organic clicks fell 38% on queries where an AI Overview appeared. Separately, Ahrefs' own aggregated Search Console data found click-through on the #1-ranked page down 34.5% by April 2025 and 58% by December 2025, comparing December 2023 to December 2025. Both measure something narrower than "all clicks" — one is organic clicks on AI-Overview queries specifically, the other is CTR for the top-ranking position specifically — which is exactly why flattening them into a round "30–40%" loses the thing that makes each figure usable.

What is good vs. bad zero-click?

Not every zero-click outcome is a loss. A buyer who asks "what's the refund window on [product]" and gets a correct, attributed answer may never click through — and didn't need to. The bad version is the same scenario with a wrong or outdated answer, no attribution at all, or a competitor cited instead of you for a question about your own category. The useful question isn't "did they click," it's "was the answer about me correct, and was I the one named." See mentions, citations and recommendations are three different things for how to separate "named" from "cited" from "recommended" in a single AI answer, since a zero-click outcome can land in any of those three states.

Who gets hit hardest: publishers or brands?

Publishers carry more direct exposure than most brands, because a publisher's business model often depends on the click itself (ad impressions, subscription conversion from an article page). A brand being described accurately in an AI answer — even with zero clicks — still delivers the thing a brand actually wants: a correct recommendation at the moment someone is deciding. That's a real distinction, not a reason to ignore the click data: a brand whose main traffic source is content marketing (guides, comparison posts, calculators) inherits the publisher's exposure, since that content earns clicks the same way a publisher's does.

What should be measured instead of clicks?

Three proxies hold up better than click count alone when clicks themselves are getting rarer:

• Named-in-answer rate: across a fixed set of buyer-style prompts, how often does an engine name your brand at all, regardless of whether anyone clicked through.
• Citation presence: when the engine does cite a source, is it your page, a competitor's, or a third party writing about the category.
• Accuracy of the claim: when you are named, is what's said about you (price, features, availability) actually correct.

How to measure AI visibility walks through building this into a repeatable check rather than a one-off glance.

What should a team actually do?

1. Stop reporting "AI Overview clicks are down" as a single number — report the specific study, its sample, and its date, every time.
2. Add a named-in-answer check to your regular reporting, even a manual one, since it survives a zero-click world better than CTR does.
3. If your content model depends heavily on click-through revenue (ads, affiliate), treat that as a distinct risk from "does AI search know who we are," since the two can move in opposite directions.

FAQ

Is "zero-click search" the same as AI Overviews? No. Zero-click search predates AI Overviews (featured snippets and knowledge panels already did this), but AI Overviews and chat-based answers have expanded how often it happens and how much text an answer can absorb before a reader needs to click anything.

Did AI Overviews cause a 30–40% drop in clicks? Not as usually stated. That exact range doesn't trace to one primary source. The real, sourced figures are actually higher: a 2026 randomized field experiment found a 38% drop in organic clicks on queries with an AI Overview, and Ahrefs' own tracking shows click-through on the #1-ranked page down 34.5% (April 2025) to 58% (December 2025). Pew's narrower per-visit click rate (8% vs. 15%) and Seer's CTR floor (0.61%) measure yet other things. "Clicks down 30–40%" looks like a rounded-down flattening of these bigger, more specific numbers, not a figure with its own source.

Does a low click-through rate mean a brand is losing business? Not necessarily. If the AI answer names and accurately describes the brand, the buyer may act without ever clicking. The risk is being absent from the answer or described wrongly, not the click itself.

Are publishers and brands affected the same way? No — publishers whose revenue depends directly on the click (ads, subscriptions) carry more exposure than a brand whose main goal is being named and described correctly, independent of a visit.

What to do next

1. Read the actual methodology section of whichever study you're about to cite — Pew, Ahrefs and Seer each disclose their own limitations plainly.
2. Build a small set of buyer-style prompts and start logging named/cited/accurate, not just rank, so you have your own proxy when clicks don't tell the story.
3. For the fuller "is this killing SEO" argument, see is SEO dead because of AI search, which works through the Pew, Ahrefs and Gartner data side by side.

ContextOwl sells one-time AI visibility reports. The sample report shows what a dated, sourced snapshot of named/cited/accurate looks like for one brand.