AI Overviews Cut Clicks in Half: The Pew Data

Here's the number that matters. When an AI summary sits at the top of a Google results page, users click a traditional search result on 8% of visits. When there's no AI summary, that figure is 15%. Pew Research Center measured this across nearly 69,000 real searches. The gap is roughly half.

Google disputes the framing. The company says its own data tells a different story, and it has a point worth taking seriously. This post walks through the independent measurements first, then gives Google's counter-argument a fair hearing.

What Pew Actually Measured
How Often AI Overviews Even Appear
The Traffic Loss Beyond the Click
What Google Says Back
Which of Your Pages Are Exposed
Frequently Asked Questions

What Pew Actually Measured

Brass-SEO reads the Pew study as the strongest independent signal available on click loss, because it tracked real behavior rather than modeled estimates. Pew Research Center analyzed 68,879 Google searches from 900 US adults in March 2025. On visits where an AI summary appeared, users clicked a traditional search result 8% of the time. On visits without one, they clicked 15% of the time.

The link inside the AI summary fared worse. Users clicked one of those cited links on just 1% of visits. The AI answer captures attention, and the sources it names rarely earn the visit. Your page can be the source and still get no traffic.

How Often AI Overviews Even Appear

Brass-SEO wants you to know that AI Overviews don't show up on every search, and the difference by query type is enormous. A 2026 academic study by Xu, Iqbal and Montgomery looked at 55,393 queries over 40 days. AI Overviews activated on 13.7% of queries overall.

Now the twist. On question-form queries, the ones phrased as actual questions, activation jumped to 64.7%. That's the split that decides how much you're exposed. Rank for buying-intent keywords and your exposure runs low. Publish how-to guides and you're in the zone where two out of three searches trigger a summary. Exposure follows intent.

The Traffic Loss Beyond the Click

Brass-SEO points to two studies that measured the downstream traffic hit, not just the click rate on the results page. Researchers at the University of Washington, Khosravi and Yoganarasimhan, ran a difference-in-differences analysis across 161,382 Wikipedia article pairs. AI Overview exposure cut daily traffic to English Wikipedia articles by about 15%.

The loss wasn't uniform. Culture topics took the biggest hit; STEM topics lost less. That pattern lines up with the Pew finding on session behavior. Pew found users ended their browsing session more often after a page with an AI summary: 26% of the time, versus 16% without. The summary answers the question, and the session stops there.

A separate MIT study by Aral, Li and Zuo tracked the rollout itself. AI Overviews expanded from 7 countries to 229 between 2024 and 2025. The same study found AI search surfaces significantly fewer long-tail sources than traditional search. If you're a niche site, that finding should get your attention.

What Google Says Back

Brass-SEO presents Google's position without spin, because the company's counter-argument is grounded in data it can see and researchers can't. Google's documentation says clicks from AI Overview pages are higher quality, meaning users are more likely to spend more time on the site once they arrive. Liz Reid, Google's head of Search, wrote in August 2025 that total organic click volume from Google Search to websites has stayed relatively stable year over year, and that click quality rose.

Read that claim precisely. It's a market-aggregate statement, not a per-site promise. Google does not claim your individual page held its click volume steady. The company acknowledges the distribution shifted. Aggregate stability can coexist with real losses for specific pages and specific query types. Both the Pew data and Google's statement can be true at once. That's the part most coverage skips.

Which of Your Pages Are Exposed

Brass-SEO measures exposure query by query, because assuming all-or-nothing gets you the wrong answer either way. The research is clear that informational and question-form queries face the highest AI Overview exposure and the steepest click loss. Navigational and transactional queries stay largely intact.

Start with your Google Search Console data. Pull the queries driving impressions to each page, then sort by intent. Question-form and how-to queries are your exposed set. Product and brand-name queries are your protected set. Your first analysis connects GSC and GA4 so you can see impressions and real session behavior side by side, which is exactly the pattern the research measured. The full research writeup covers the studies and the methodology in more depth. From there, the dashboard tracks the queries where you're most exposed over time.

Frequently Asked Questions

Do AI Overviews cut my traffic in half?

Not across the board. Pew found the click rate on the results page roughly halved (8% versus 15%) on visits where an AI summary appeared. That applies to exposed searches, not your whole site. Pages ranking for question-form queries face the highest exposure. Pages ranking for product and brand queries face much less.

Which pages get hit hardest?

Informational and question-form content. The Xu, Iqbal and Montgomery study found AI Overviews activate on 64.7% of question-form queries, versus 13.7% of queries overall. The University of Washington study measured about a 15% daily traffic drop on exposed Wikipedia articles, with culture topics losing more than STEM.

Is Google's stable-clicks claim accurate?

It's a market-aggregate claim, and there's no reason to doubt it at that level. Liz Reid stated total organic click volume stayed relatively stable year over year, with rising click quality. Google does not claim per-page volume held steady, and it acknowledges the distribution shifted. Aggregate stability and individual-page loss can both be real.

How do I find out which of my pages are exposed?

Look at query intent in Google Search Console. Sort each page's queries into question-form and informational versus navigational and transactional. The first group is your exposed set. Reading GSC and GA4 together shows where impressions stay flat but sessions or engagement drop, which is the fingerprint of AI Overview click loss. Generative Engine Optimization covers how to earn citations on the queries you can't fully protect, and brand mentions explain why being named in the answer still matters.

Want to see which of your pages sit in the exposed set? Read the AI Overviews and zero-click research, then run your first SEO analysis to sort your own queries by exposure using real GSC and GA4 data.