Google Confirms Disabled Discover Source Clicks Were a Bug in an Experiment

Google has confirmed that disabled source taps on Discover's AI summary cards were "a bug in a Discover experiment," and says it has already switched the experiment off. Rajan Patel, VP of Engineering for Search at Google, gave that confirmation in a reply on X Monday evening, closing out a four-day flap that began with an independent researcher's report on October 23.

The bug sat at a sensitive spot: the only path from Discover's AI summaries to the publisher sites they cite.

What exactly was broken?

Damien Andell, who researches Google Discover, reported on October 23 that Discover was testing an AI summary change that disabled clicks on sources. He found the test through a flag called "disable_aio_sources_button_tap" — a server-side switch that controls which users see a given feature. That flag, Andell stated, stopped taps on the sources at the bottom of AI summary cards from opening the article feed.

Before the test, tapping a summary's sources opened a feed of articles from those cited sources. That feed is one of the few ways a user can get from an AI summary in Discover to the sites it draws from.

Andell's evidence came in two screenshots of his own Discover feed. Both show a Fox 9 video from YouTube with a brief summary underneath and a row of icon links below that. His second image circles those icons and labels them "Sources."

What did Google actually say?

Patel wasn't replying to Andell directly. He was responding to Lily Ray, who had quote-posted Andell's report on the morning of October 23. Patel wrote:

Hey Lily – this was a bug in a Discover experiment. We've turned off the experiment, and the experience works as intended again.

That's the full extent of Google's statement, and it leaves three questions open. The reply doesn't say what the experiment was testing, how long it ran, or how many users saw the disabled taps.

It's worth separating what's confirmed from what isn't. Confirmed by Google: the disabled taps were a bug, the experiment is off, and the sources experience works again. Everything else — the experiment's purpose, its reach, its duration — remains tool-signal territory, based on Andell's flag-watching rather than any Google statement.

What happened after Google's reply?

The experiment didn't simply disappear. On October 27, the morning after Patel's reply, Andell checked again and posted on X:

I just checked and the experiment did indeed remove and replace it with another one called "aio tap falls through to hero".

His post didn't explain how the new experiment works or where it is running. The name suggests taps on sources may now route to a card's hero content rather than an article feed, but that's inference from the flag's name — Google has said nothing about it.

Why publishers should care

For site owners, this touches the thin reporting surface Google currently offers for Discover's AI features. The Discover generative AI report in Search Console (Google's performance dashboard for search and Discover) counts how many times users see links to your pages in Discover through generative AI capabilities, with each source page listed by URL.

There's a catch, though. The only metric Search Console Help describes for that report is Impression — when someone sees one of your site's links via generative AI in Discover. It does not count source taps. So even before this bug, publishers had no direct measurement of clicks from AI summary cards, only visibility.

What to watch next

Until Google or Andell explains the replacement experiment, you can check it yourself: tap the sources on an AI summary card in your own Discover feed and see whether the article feed opens. Watch for any follow-up flag discoveries from Andell, and for whether Google ever adds a tap or click metric to the Discover generative AI report.