Google says disabled source taps on Google Discover AI summary cards were a bug in a Discover experiment, Search Engine Journal reports. Rajan Patel, VP of Engineering for Search at Google, said so in a reply on X on Monday evening, and said the experiment has been turned off. Damien Andell, who first reported the issue, says it has since been replaced with another experiment.
What happened to Google Discover source taps?
Damien Andell, a researcher who tracks Google Discover, reported on Oct. 23 that Discover was testing an AI summary change that disabled taps on sources. He found the test through an internal flag called disable_aio_sources_button_tap (an internal experiment configuration key that controlled how taps on the source icons behaved). Before the experiment, tapping a source on a Discover AI summary card opened a feed of articles from those sources, one path from an AI-generated summary to the websites it cited, Andell wrote.
Lily Ray had quote-posted Andell’s report on the morning of Oct. 23. Patel replied on X on Monday evening, Oct. 26:
“Hey Lily, this was a bug in a Discover experiment. We’ve turned off the experiment, and the experience works as intended again.”
Patel’s reply did not say what the experiment was originally testing, how long it ran, or how many users encountered the disabled taps, per Search Engine Journal.
What experiment replaced the original one?
The morning after Patel’s reply, Andell found the original experiment had been removed and replaced with a new one. In a post on X, he stated: “I just checked and the experiment did indeed remove and replace it with another one called ‘aio tap falls through to hero’.” He did not explain how the new experiment works or where it is running, according to Search Engine Journal.
What does the Search Console Generative AI report actually measure?
Google’s Generative AI performance report for Discover tracks one metric: Impressions. An impression is counted when a user sees a link to your site through generative AI features in Discover. According to Google Search Console Help, the report was rolled out to all websites worldwide as of August 31, 2026.
Source taps are not a tracked metric in this report. Google Search Console Help also notes that the report does not include data from experiments in Search Labs (Google’s program for testing early-stage features with opted-in users), as those are still in active development. Whether the Discover experiment Andell identified falls under Search Labs is not specified in either source.
What this means for AI-search visibility
If your content appears in a Google Discover AI summary, you can see that it was shown. You cannot see whether anyone actually tapped through to read it.
Picture it concretely. Your article is listed as a source under a Discover AI card. Search Console logs an impression. But Search Console has no tap or click-through metric for this report. During the window this bug was active, taps were disabled for affected users, and the report has no metric that would have shown it.
In Hingewise’s view, the replacement experiment is the more pressing open question. The name “aio tap falls through to hero” suggests that tapping a source might now route users to the primary content item in the card (the “hero” element, likely the main article or video at the top of the card) rather than opening a dedicated feed of source articles. If that is how it works, it would reduce the direct path from an AI summary to the individual publisher sites listed as sources. Neither Google nor Andell has explained the mechanics publicly.
The structural gap is clear: as Google iterates on how Discover AI summaries behave, publishers are working with impressions as the sole signal in Search Console. Impressions show reach. They do not show whether that reach translates to visits. Until Google adds a tap or click metric to the Generative AI performance report, publishers cannot measure the referral value of being cited in Google Discover AI summaries from Search Console alone.
What to check before the next experiment ships
- Open Search Console and pull your Generative AI performance report for Discover. Review impression trends across the Oct. 23–27 window.
- Keep in mind the report measures impressions only, not taps or click-throughs to your site.
- Test source taps in your own Discover feed to observe how the current replacement experiment behaves.
- Cross-reference Discover referral traffic in your analytics for that same window to see if a drop aligns with the bug period.
- Monitor Damien Andell’s posts on X for further analysis of the “aio tap falls through to hero” experiment.
- Do not rely on Search Console alone to alert you if source-tap behavior changes silently in a future experiment.
Sources: Search Engine Journal, Oct. 27, 2026; Google Search Console Help, Generative AI performance report (Discover)
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.