Google Blames Disabled Discover Source Clicks on Experiment Bug
The Development
Rajan Patel, a Google executive, wrote on X that the disabled source clicks on Google Discover AI summaries were a bug in an experiment, and that the experiment has been turned off. Publishers had flagged AI-generated summaries in Discover where the cited sources could not be clicked, cutting off referral traffic from a surface that sends significant volume to news and content sites. Patel's post is the only confirmation so far; Google has not published details on the experiment's scope, how many users saw it, or how long it ran. Search Engine Journal reported the statement on October 27, 2026.
Our Take
Take the bug explanation at face value and the strategic read stays the same. Google built, shipped, and tested a Discover variant where the summary replaces the source, and the failure was caught only after publishers noticed. Discover is the one Google surface where publishers get traffic without a query, so a no-click version removes their last discovery channel that does not depend on ranking against AI answers. The pattern matches AI Overviews in Search: ship the summary layer, adjust link treatment under pressure. Expect link prominence in Discover to remain a recurring negotiation, not a settled feature.
What Changed
Google can now generate AI summaries inside Discover, a feed with no search query, and test them live with source attribution selectively disabled. Discover is no longer only a referral surface; it is a place where Google can deliver the answer without the click.
Marketing Impact
Content and SEO teams that count Discover as a stable referral line need a contingency in their traffic forecasts. Performance teams relying on Discover for top-of-funnel content distribution should model a scenario where AI summaries cut click-through. Brand teams should track whether their content is cited inside summaries, not just whether it ranks.
Competitive Implication
Publishers with direct audience relationships, such as newsletters, apps, and logged-in users, gain relative strength because they depend less on Discover referral. Publishers whose traffic skews toward Discover and search referral are most exposed, since Google controls both the summary layer and the link treatment, and can change either without notice.
Strategic Outlook
Google will test AI summaries in Discover again, with source links likely retained but demoted in prominence. Expect more publisher-reported anomalies surfacing through social posts before any official documentation. Discover referral volatility will rise through Q1 2027, and publishers will push for traffic transparency, as they have with AI Overviews.
The Exploit
Action Item
Heads of content and audience at publishers and content-led brands should pull 90 days of Discover referral data by Friday and set a variance alert at a 20% week-over-week drop, so the next experiment is detected in days, not weeks.