Discover's New "Hero" Flag Follows a Google Bug Reply

You read a Google "bug" reply as the end of a Discover test. This one may be the start of the next.

On October 23, 2026, Discover watcher Damien Andell posted that Google was testing AI cards where tapping the sources did nothing. By his account the flag was named disable_aio_sources_button_tap. On October 27, Google's Rajan Patel said it was a bug and the experiment was off. Andell then reported that a flag called "aio tap falls through to hero" had replaced it.

Nobody has said which source that tap would favor. For publishers that is the question to watch.

Search Engine Roundtable broke the story on October 26, and its reporting is where the timeline starts.

Andell wrote that before the test, tapping the sources at the bottom of an AI summary card opened a mini-feed of articles from those sources. With the test, he said, the flag switched that off. A Discover card is one of the few places where the AI summary and the publisher links sit side by side. Readers get the summary and no link onward.

Publishers reacted fast. Andrej, who runs severe-weather.eu, replied under Andell's post: "Wow goodbye Discover traffic, I guess". Schwartz, who runs Search Engine Roundtable, wrote that he tried to replicate the test and could not. At that point the evidence was Andell's post and little else.

The reply from Google came on October 27. Rajan Patel of Google answered a post from SEO Lily Ray on X. "Hey Lily - this was a bug in a Discover experiment. We've turned off the experiment, and the experience works as intended again," he wrote.

Ray thanked Google for reinstating links to publishers in Discover. Schwartz wrote on X, asking whether it was "a bug or an experiment or both?" and added Google's answer to his story.

The flag that came next

Andell then checked the app again. In a reply he wrote on X that the experiment "did indeed remove and replace it with another one called "aio tap falls through to hero"." Schwartz carried the same line in his update. The name is all anyone has. Andell has not said what the new flag does on screen. I won't guess at the behavior.

The names do not settle whether the first test was a bug or a plan, and I can't tell from a flag. My read is that a setting called disable reads like something somebody chose to build, whatever Google decided afterward. That is an opinion, and Google's own account is the one above.

This is also not the first time Discover's AI cards have been tuned in ways that change what publishers get. Google added AI summaries to Discover in 2025. Andell has said that in October 2025 a separate test removed the "Generated with AI" label, using a flag called "disable AI summary disclaimer". Each change was small. Together they show a surface that Google keeps adjusting, mostly out of view.

Andell also said on October 23 that tests like this often never ship or don't last. I agree that a flag is not a launch. He added that Google may be seeing low click-through on the sources, which is his guess and not Google's explanation.

What a publisher can check

The window from October 23 to 27 is real, and it is the kind of thing that wrecks a traffic report. Open the Discover report in Search Console and look at those days. If impressions held and click-through rate fell, an experiment is a likelier explanation than your content. Annotate the dates before anyone rewrites a headline to fix a problem a product test caused. I would diagnose first and prescribe second, and Discover makes that hard because the report shows results and almost never the cause.

I think "hero" matters most here. If a tap on an AI card ends up opening one article, one publisher gets the visit and the rest of the sources lose theirs. Google has not said how it would pick. Until it does, I'd treat October 27 as a pause in the testing and not the end of it.