Google Search's product chief will discuss trust at scale. Clicks: 8% vs 15%

TechCrunch has announced that Robby Stein, vice president of product for Google Search, will speak at Disrupt in October on how product decisions change at scale. The post is conference marketing with a ticket deadline attached, so the announcement is not the news. The job title is. Stein runs product for the system that decides whether anyone reaches a publisher at all, during the years it started answering questions itself.

What that change did to the other side of the arrangement has been measured, now that an LLM writes the answer above the links. Pew Research Center went through roughly 70,000 searches from about 900 users and compared searches with an AI answer to searches without one:

• A result is clicked on 8% of searches with an AI answer, against 15% without.
• Links inside the AI answer itself are clicked about 1% of the time.
• The session ends right there on 26% of AI-answer searches, against 16% without.

Read those three lines together. The click rate falls by 46.7% in relative terms when an AI answer is present, the citations inside the answer are clicked about an eighth as often as an ordinary result, and the share of searches that end the browsing session outright rises by 10 percentage points, which is 62.5% more often. The answer is not competing with the links. It is replacing the visit.

The question in the session description

“How do you introduce something new without undermining what people already value?”

— TechCrunch, From the Disrupt 2026 session description

TechCrunch, describing the session Stein will lead

That question is not rhetorical for anyone who publishes. It has an arithmetic answer now, and the answer is 8% against 15%. Stein's session is framed around balancing speed with trust and innovation with reliability, which is the vocabulary of a product team looking inward at its own users. The people on the other side of the result page are not users of Google. They are the supply.

What a publisher does about it

We say this as a site that lives on search referrals, so the interest is declared. The practical response is not to complain about the numbers but to build the kind of material that survives them: reference pages people return to directly, not answers a summary can absorb in one line. That is the reasoning behind our asset guides covering 83% of the market.

And there is a pattern worth naming across this week. A feature is announced, the world is rearranged around it, and the measurement arrives later and from outside the company that shipped it. Apple has just paid $250m over an AI feature it announced and delivered 813 days later. In search the feature shipped on time. It was the accounting of who pays for it that ran behind.

Informational material, not investment advice. The click figures are from Pew Research Center's analysis of roughly 70,000 searches by about 900 users; the announcement itself is conference marketing.