Google has said that an issue preventing users from opening cited-source content from some Discover AI summaries was a bug in an experiment, not a permanent change to how Discover links to publishers.
The episode matters because the source area in an AI-generated summary can be a route from Discover to the articles supporting it. If that route does not respond to a tap, a publisher may still appear in the experience but lose an opportunity for a reader to reach its content. The available evidence points to a temporary and limited UX experiment, rather than a broadly announced Discover product update.
On October 23, Discover observer Damien Andell posted that a flag called disable_aio_sources_button_tap was active in Discover. According to his observation, tapping the Sources area on an AI summary did not open the associated article feed.
Google engineer Rajan Patel subsequently responded that the behavior was a bug in a Discover experiment. He said the experiment had been turned off and that the experience was "as intended again." Search Engine Journal's reporting on the sequence links the posts and describes the events between October 23 and 27.
That clarification is important. Google confirmed the experimental bug and its shutdown, but it did not publish a detailed account of the experiment's audience, geographic scope, duration, or the number of users affected. There is therefore no basis to treat the incident as a universal Discover traffic change.
In the days after the issue was reported, Andell identified what appeared to be another test, described as "aio tap falls through to hero." That wording suggests a potentially different tap-routing behavior, but Google has not publicly documented the flow or confirmed what it means for source links. It should be treated as an observed signal, not as an announced feature specification.
Discover AI summaries can present generated content alongside source material. A publisher's presence in that interface and a reader's ability to reach the publisher are related, but they are not the same thing.
When a UI change alters where a tap goes, it can affect the path between an AI summary and an article feed even if the publisher remains represented in the summary. For editorial teams, this is a useful reminder that visibility metrics do not automatically describe every downstream interaction.
The likely practical concern during an incident like this is referral access, not indexing or eligibility. Nothing in the available reporting indicates that the experimental bug changed whether pages could be crawled, indexed, or selected for Discover. Nor does it establish a measurable traffic impact for any individual publisher.
Google's Discover Generative AI performance report documentation confirms that Search Console includes reporting for impressions from Discover Generative AI features. This gives publishers a way to monitor exposure in the relevant Discover experience.
However, the documentation does not describe per-interface tap events, source-panel interactions, or the specific experiments reported in October. That creates an important measurement boundary:
Publishers reviewing the period should avoid assuming that stable or rising Generative AI impressions prove that source-driven visits were unaffected. Equally, a referral change cannot be attributed to this experiment without corroborating evidence, because Discover traffic naturally varies with audience interest, content timing, and other ranking and product factors.
Google's broader Discover guidance remains the appropriate reference for the service's documented content and appearance considerations. The source-tap incident does not add a new published optimization requirement.
The short duration and uncertain scope mean this is not a reason to overhaul a Discover strategy. It is a reason to maintain enough measurement discipline to spot meaningful changes when Google tests AI-driven interfaces.
For publishers with Discover exposure, useful checks include: The key distinction is between a potential change in user interaction and a confirmed change in distribution. The evidence supports the former during an experiment. It does not support claims of a lasting ranking, indexing, or reporting change.
For teams trying to understand how their content is represented as search interfaces evolve, create a GEO Search Leads account to assess your website's GEO readiness and identify optimization opportunities. That can help turn questions about AI-generated discovery into a structured review of content accessibility, entity clarity, and areas that may deserve closer monitoring. Did Google permanently disable links to sources in Discover AI summaries?
No. Google engineer Rajan Patel said the disabled source-tap behavior was a bug in a Discover experiment and that the experiment had been turned off. Public reporting does not establish a permanent removal of source access.
Did the experiment affect every Discover user?
That is not known. Google did not publicly provide the experiment's reach, locations, devices, or duration. The available evidence describes an observed experiment rather than a global product change.
Will the Discover Generative AI report show source taps or clicks?
Google's documentation confirms the report tracks impressions for Discover Generative AI features. It does not document reporting for individual source taps or the routing of taps within an AI summary.
Should publishers change their Discover SEO strategy because of this bug?
There is no evidence that the bug changed Discover eligibility, crawling, indexing, or ranking requirements. Publishers should monitor relevant traffic and impression trends, but a broad strategy change is not supported by the available information.
Google's response indicates that the disabled source-tap behavior was an experimental bug that was switched off. For publishers, the more durable lesson is measurement: AI-generated Discover visibility and the paths that lead readers to source content may not always be captured by the same metric. Monitor both carefully, and treat isolated UX-test signals with appropriate caution.