Do AI Overviews reduce clicks to my website?

What the click data shows#

On searches where an AI Overview appears, fewer people click through to any website. Pew Research followed about 900 US adults through 68,879 real Google searches in March 2025. When an AI summary was on the page, visitors clicked a traditional search result on 8% of visits. Without a summary, that number was 15%, nearly double. The links inside the summary itself fared worse: users clicked one on just 1% of visits.

The effect goes beyond the click. With an AI summary present, people ended their session right there on the results page 26% of the time, compared with 16% when no summary appeared. The answer was complete enough that they had no reason to go anywhere.

A common hope is that AI referral traffic quietly replaces those lost search clicks. It does not, at least not yet. Across major industries in late 2025, AI assistants drove only about 1% of overall web traffic (Conductor data, reported by Digiday). The clicks leaving through the front door are far larger than the trickle coming in from ChatGPT and its peers.

Why the citation matters more than the click#

When the answer names you, you win the buyer who never clicks anything. That is the shift AI Overviews force. The visibility used to be the blue link and the visit it produced. Now a large share of the value sits inside the summary, in whether your brand is the one the AI cites and recommends.

This changes what you optimize for:

• Measure presence in the answer, not just sessions. A dashboard built on referral clicks will show AI search as a rounding error and tell you to ignore it, right up until it has shaped who your buyers shortlist.
• Earn the citation, not the ranking. Being mentioned in the summary is the impression that does the persuading. It travels with the answer whether or not anyone taps through.
• Treat the remaining clicks as higher intent. When the summary answers the simple lookups, the buyers who still click through are self-selecting for a deeper need, so the visits that survive an AI Overview skew more engaged. Fewer, warmer clicks is the pattern to plan around.

Here is the one play to run:

1. List the buyer queries in your category where an AI Overview already appears (search them and look).
2. For each, check whether your site is one of the linked sources inside the summary. Prefer checks this for you.
3. Where it is not, tighten the page the Overview pulls from: answer that exact question in the first line, keep it crawlable, and earn corroboration on the third-party sources it cites.

You’ll know it worked when your brand starts showing up as a linked source inside those AI Overviews, and the share of your tracked queries where you are cited climbs run over run, even as raw organic clicks stay flat.

None of this means clicks stop mattering. It means the scoreboard changed. If you only count visits, you will conclude AI search barely exists. If you count how often you are the cited source, you will see the real story. Our four-engine citation study shows which sources engines actually quote, and our guide to appearing in Google AI Overviews covers how to become one of them.

To see how often AI engines already name your brand across ChatGPT, Gemini, Perplexity, and Google AI Overviews, run a free AI visibility audit.

Sources

1. Pew Research tracked 68,879 Google searches by about 900 US adults in March 2025; roughly 18% produced an AI summary. On those, 8% of visits clicked a traditional result versus 15% without a summary, and only 1% clicked a link inside the summary. Pew Research Center, March 2025 browsing data
2. In the same Pew data, users ended their browsing session on the results page 26% of the time when an AI summary was present, versus 16% without one. Pew Research Center, March 2025 browsing data
3. AI assistants drove only about 1% of overall web traffic across major industries in late 2025 (Conductor data reported by Digiday), so referral clicks from AI tools do not offset the search clicks lost to summaries. Digiday, state of AI referral traffic 2025 (Conductor data)