Google says the disabled source taps on Discover's AI summary cards were a bug in an experiment, now switched off — the one click path from a summary to the sites it cites briefly stopped working.
2026-10-29 · by Moe Ameen
On October 27, 2026, Google said the disabled source taps spotted on Google Discover's AI summary cards were "a bug in a Discover experiment," and that it had turned the experiment off. Rajan Patel, VP of Engineering for Search at Google, posted the reply on X, adding that "the experience works as intended again."
The issue was first flagged on October 23 by Damien Andell, who researches Google Discover. He traced it to an experiment flag named "disable_aio_sources_button_tap" that stopped taps on the sources row at the bottom of an AI summary card from doing anything. Normally, tapping those source icons opens a mini-feed of articles from the sites the summary cites — one of the few paths that sends a reader from an AI summary to a publisher's page. Under the experiment, those taps were dead. Search Engine Roundtable's Barry Schwartz tried to reproduce it on October 26 and could not, which is consistent with a limited, flag-gated test rather than a broad rollout.
Google's reply did not say what the experiment was testing, how long it ran, how many people saw the disabled taps, or where it ran. Separately, Andell reported on October 27 that the experiment had been removed and replaced with a different one, labeled "aio tap falls through to hero" — Google has not explained how that behaves.
The episode sits inside a longer-running tension. Google added AI summaries to Discover in 2025, a move publishers objected to because a summary that answers in the feed gives readers less reason to tap through. A bug that silently removes even the source-tap path — the thin remaining bridge to the cited site — is the kind of change publishers have no visibility into and no control over, whether it is intentional or, as here, accidental.
This is a small bug with a big lesson: the bridge from a Google AI summary to your actual page is a single tap that Google can disable with a flag, and did — by accident, for days, with no disclosure. Discover referral traffic is real and worth earning, but it is a borrowed channel. The defense is not to chase the summary card harder; it is to stop letting one algorithm-controlled surface be the only place your audience can find you. That means publishing your own content, in your own voice, directly to the platforms where people already follow you — so a reader who never taps a source link still sees your work.
[Kompozy](/) is built for exactly that distribution breadth. Point it at your reporting, your video, or a topic and it generates native posts across the publishing surface — eight social platforms plus blog and email — governed by your [Persona Brief](/glossary/persona-brief) so each one reads like you wrote it for that feed, not a syndicated dump. A long piece becomes a [Clipped Short](/glossary/content-repurposing) for Reels and Shorts, a [Carousel](/glossary/content-repurposing) for Instagram and LinkedIn, a Text Post for X, and an [Email Newsletter](/glossary/content-repurposing) to the list you own outright — then [Autopilot](/glossary/autopilot) schedules and fans them out on a cadence. When Google changes how Discover taps behave, the audience you reached directly is untouched. For the strategy side, pair this with our guides on [content strategy beyond Google traffic](/guides/content-strategy-beyond-google-traffic) and [getting your content cited in AI Overviews](/how-to/get-your-content-cited-in-google-ai-overviews).
A Discover experiment, gated by a flag named "disable_aio_sources_button_tap," disabled taps on the sources row at the bottom of an AI summary card. Normally those taps open a feed of articles from the sites the summary cites, so while the experiment was live, that click path to cited publishers stopped working. Google said on October 27, 2026 that it was a bug and turned the experiment off.
No. Rajan Patel, VP of Engineering for Search at Google, said only that it was "a bug in a Discover experiment," that the experiment had been turned off, and that the experience works as intended again. Google did not disclose what the experiment was testing, how long it ran, how many people saw the disabled taps, or where it ran.
It shows the referral path from a Discover AI summary to your page is thin and controlled entirely by Google — a single experiment flag disabled it for days before an outside researcher noticed. Being cited as a source does not guarantee a visit, and the behavior can change without notice. The practical response is to diversify distribution onto channels you own, rather than depending on an AI summary card for traffic.
Kompozy generates native posts across eight social platforms plus blog and email from one source, each adapted to the feed and written in your brand voice via the Persona Brief, then schedules and publishes them on autopilot. That builds reach on channels you control directly, so a change to how Discover handles AI summary taps does not decide whether your audience sees your work.