Google Attributes Disabled Discover Source Clicks to a Bug in an AI Summary Experiment

By Matt G. Southern | Enriched & Expanded Report

Executive Summary: The Main Facts

Google has officially clarified that recent user-facing friction regarding Google Discover—specifically, the inability to tap or click on source citation cards generated by artificial intelligence—was entirely unintentional. According to Rajan Patel, Google’s Vice President of Engineering for Search, the disabled navigation resulted from an unexpected software bug tied to a routine Discover UI experiment.

The issue first came to light in late October when search engine researchers and digital marketers noticed that interactive source buttons embedded beneath AI-generated summary cards within the Google Discover feed had lost their functionality. Instead of redirecting users to a deeper feed of related content or cited articles, tapping these sources yielded no response.

Following public inquiries on social media platforms by prominent SEO professionals, Google intervened, disabled the problematic test, and restored the original user experience. However, the incident has reignited broader industry conversations regarding how Google tests AI-driven features in its ecosystem, the transparency of algorithmic experiments, and the critical importance of outbound traffic to publishers who rely on Google Discover for referral visibility.

Chronology of Events: How the Bug Unfolded

The discovery, tracking, and resolution of the Google Discover citation bug unfolded rapidly over the course of several days across late October, primarily documented by search researchers on X (formerly Twitter).

October 23: The Initial Discovery

Search researcher Damien Andell, who closely monitors updates and hidden features within Google Discover, first flagged the anomaly. Andell discovered that Google was rolling out an AI summary iteration that effectively blocked outbound clicks on source icons.

By inspecting application flags, Andell identified a specific system variable designated as disable_aio_sources_button_tap. This internal flag was tied to an ongoing experiment governing Google’s Search Generative Experience (SGE) style summaries within the Discover feed. In his public reports, Andell shared visual evidence, including screenshots of a YouTube-hosted video from Fox 9 appearing in his Discover feed, accompanied by an AI-generated summary. Below the text block, a row of publisher and platform icons was labeled "Sources," but interacting with them failed to trigger the expected article feed.

October 23 (Later That Day): SEO Community Amplification

Recognizing the potential traffic ramifications for publishers, prominent SEO strategist Lily Ray quote-posted Andell’s findings on X, drawing broader industry attention to the issue. The realization that AI-driven summaries might sever the direct user path to original publishers caused immediate concern within the SEO and publishing communities, where Google Discover is often a massive driver of volatile yet high-volume referral traffic.

October 27: Google’s Intervention and Resolution

Over the weekend following the initial reports, Google engineers investigated the codebase. Late Monday evening, October 27, Rajan Patel formally responded to Lily Ray’s thread. Patel explicitly stated that the broken functionality was not a permanent product shift, but rather an isolated bug within a specific trial run.

By the morning of October 28, Andell returned to the platform to confirm that Google had indeed pulled the plug on the malfunctioning trial. However, his follow-up analysis revealed that the experiment had been quickly swapped out for a different iteration carrying the internal flag label aio tap falls through to hero. As of yet, Google has not provided detailed documentation on what this subsequent experiment aims to test.

Supporting Data and Technical Context

To understand why this bug caused such immediate alarm, one must examine how Google Discover integrates artificial intelligence summaries and how those interactions are tracked by webmasters.

The Anatomy of a Discover AI Summary Card

Google Discover relies heavily on machine learning to serve personalized content streams to mobile users. In recent months, these feeds have increasingly incorporated AI-generated summaries—short, synthesized overviews designed to give users a quick snapshot of a topic or video before they commit to reading the full article.

Under normal operating parameters, these summary cards display a distinct row of source citations at the bottom. Tapping these citations historically opened a curated feed of articles from those specific publishers, serving as a vital bridge between Google’s AI-generated content layer and the original creators. The experiment identified by Andell effectively severed this bridge by short-circuiting the button-tap event listeners.

Search Console and the Discover Generative AI Report

For publishers, visibility within Google Discover is monitored via Search Console’s dedicated Discover Generative AI Report. This specialized analytics dashboard tracks how frequently user sessions encounter website links generated through Discover’s AI features.

However, the reporting metrics associated with this feature remain limited. According to official Search Console documentation, the primary metric captured in this report is Impressions—defined strictly as the number of times a website’s URL is displayed via Discover’s AI summaries. It does not natively break down interactive user engagements, such as active source taps or downstream click-through rates (CTR). Consequently, if a bug disables user taps, publishers face a blind spot where impressions might continue to register, but actual traffic acquisition drops to zero without an immediate explanation in standard analytics dashboards.

Official Responses from Google Leadership

Google’s response to the incident was swift, though characteristically concise.

Responding directly to Lily Ray on X, Rajan Patel wrote:

"Hey Lily — this was a bug in a Discover experiment. We’ve turned off the experiment, and the experience works as intended again."

While Patel’s statement successfully quieted fears of a permanent UI overhaul, it left several technical questions unanswered. Google’s communications did not disclose:

1. The Scope of the Test: Exactly how many global users were exposed to the disabled source button bug during the active experiment window.
2. The Purpose of the Flag: What user behavior or metric the disable_aio_sources_button_tap experiment was actually trying to evaluate.
3. The Timeline: How long the buggy version of the experiment had been live prior to Andell’s initial October 23 discovery.

Despite the lack of granular data, the speed at which Google acknowledged and disabled the experiment reflects heightened sensitivity toward publisher relations, particularly concerning AI search features that aggregate third-party content.

Implications for SEOs and Digital Publishers

While Google has confirmed that the bug is resolved and normal functionality has been restored, the episode highlights critical underlying challenges for the digital publishing ecosystem.

1. The Vulnerability of AI-Mediated Traffic

As Google continues to weave generative artificial intelligence deeper into surfaces like Search and Discover, the line between content creation and content aggregation blurs. When an AI summary card answers a user’s query or summarizes a video directly on the feed page, the incentive to click through to a publisher’s website diminishes. A bug that actively disables citation links demonstrates just how fragile publisher referral pipelines can be when intermediated by automated UI experiments.

2. The Critical Role of Independent Researchers

This incident underscores the immense value of independent search researchers—such as Damien Andell and various SEO analysts—who manually audit Google’s interface changes. Because major tech platforms frequently push out silent, A/B-tested UI tweaks without prior public announcements, the digital marketing community relies heavily on reverse-engineering and continuous monitoring to catch disruptions early.

3. Analytics Transparency Needs Improvement

The disconnect between Search Console’s current reporting capabilities and the realities of AI-driven feeds remains a persistent pain point. Because the Discover Generative AI Report measures impressions rather than deep user engagement metrics, publishers are often left guessing whether flat traffic trends are the result of poor content alignment, algorithmic shifts, or underlying UI bugs like the one recently patched by Google.

Looking Ahead

As Google continues to iterate on its AI infrastructure across Search and Discover, webmasters and SEO professionals are advised to maintain active monitoring protocols.

If you suspect anomalous traffic drops or interact with your own mobile Discover feeds, simple manual audits—such as tapping the source attribution cards on AI summary blocks—remain an effective frontline diagnostic tool. Until Google expands its Search Console reporting to offer clearer attribution metrics for AI-driven summaries, vigilance and community collaboration will remain the publisher’s best defense against unexpected algorithmic and interface anomalies.