Google clarifies that disabled source clicks in Discover AI summaries were a bug in an experiment, but new tests suggest ongoing changes to user traffic.

The delicate relationship between AI-driven summaries and original content creators reached a point of tension this week when users noticed a critical failure in Google Discover. For publishers who rely on the platform for a significant portion of their referral traffic, the discovery that source links within AI summary cards were becoming unresponsive was more than a technical glitch; it felt like a strategic erasure.

The issue first came to light through the observations of Damien Andell, a researcher specializing in Google Discover. Andell identified a specific code flag, disable_aio_sources_button_tap, which effectively neutralized the sources button at the bottom of AI-generated summary cards. Under normal operating conditions, tapping this button opens a mini-feed containing the specific articles Google used to synthesize the AI response, providing a direct pathway for the user to visit the original publisher.

The anatomy of a disappearing click

When the experiment was active, the functionality that usually bridges the gap between an AI answer and a source website vanished. In documented cases, such as a Fox 9 video appearing in the Discover feed, the AI summary remained visible, and the source icons were present, but they ceased to function as hyperlinks. This created a dead-end user experience where the AI provided the value, but the creator of the underlying data was stripped of the visit.

This specific mechanism is vital because Google integrated AI summaries into Discover in 2025. At the time, the publishing industry expressed widespread concern that generative AI would satisfy user curiosity directly on the search page, removing the incentive to click through to the source. The sources button was viewed as one of the few remaining safeguards ensuring that high-quality reporting still translated into tangible web traffic.

Google responds to the outcry

As the news spread through the SEO community and was amplified by experts like Lily Ray, Google was forced to provide a clarification. Rajan Patel, the VP of Engineering for Search at Google, addressed the situation on X, stating that the disabled source taps were the result of a bug within a Discover experiment. According to Patel, the experiment in question has been turned off, and the user experience is now working as intended.

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

While the official stance labels the event as a bug, the lack of detail regarding the experiment's original purpose—or how many users were affected—has left some digital marketers skeptical. The transition from a functional link to a disabled button is a significant behavioral change, and in the world of algorithmic testing, the line between a bug and a test of a new "feature" is often thin.

A new experiment emerges immediately

The resolution of the bug did not bring a return to the status quo. Shortly after Patel's confirmation that the faulty experiment was deactivated, Damien Andell discovered that it had been replaced by another test. This new iteration is identified by the flag aio tap falls through to hero.

Although the specific mechanics of this new test remain unclear, its immediate deployment suggests that Google is still aggressively iterating on how AI summaries interact with the Discover interface. For entrepreneurs and media owners, this indicates that the "intended experience" is a moving target. The shift from disabling clicks to a "fall through" mechanism suggests Google is searching for a way to balance AI utility with the necessity of sending traffic to the web ecosystem.

The measurement gap in Search Console

One of the most frustrating aspects of this volatility for business owners is the lack of granular data. Currently, the Discover generative AI report in Google Search Console provides a limited view of performance. The primary metric available is Impressions, which tracks how many times a link to a page is seen within a generative AI response.

Crucially, the report does not track "source taps" as a distinct metric. This means that when a bug or an experiment disables clicks, publishers may see their impressions remain steady or even increase—because the AI is still citing them—while their actual click-through rate (CTR) plummets. This data gap makes it difficult for businesses to prove traffic loss in real-time unless they are monitoring their own server logs with extreme precision.

Strategic implications for digital publishers

The recent instability in Discover highlights a broader trend in the AI era: the transition from a "link-based" economy to an "answer-based" economy. When Google tests the removal of source clicks, it is essentially testing the viability of a closed-loop system where the user never leaves the Google ecosystem.

To mitigate these risks, businesses are increasingly diversifying their traffic sources. The reliance on a single algorithmic feed for discovery is becoming a liability. Strategies now include:

Building direct-to-consumer channels via newsletters and proprietary apps to bypass the "AI middleman."

Focusing on high-intent, long-form content that AI cannot easily summarize in a few bullet points.

Optimizing for "brand mentions" rather than just "clicks," recognizing that an impression in an AI summary still carries some brand equity, even if the click is missing.

Global market outlook and business impact

For companies operating in the USA, UK, and global markets, this incident underscores the precarious nature of the "AI-Publisher" pact. In the United States, where the legal landscape regarding AI training and fair use is still being litigated in courts, any move by Google to intentionally restrict traffic to sources could be viewed as an exacerbation of the tension between tech giants and content creators.

In the UK and other global markets, where digital competition laws are becoming more stringent, the transparency of these "experiments" is under scrutiny. If a dominant platform can accidentally or intentionally disable the primary revenue driver for publishers (traffic) via a "bug," it raises questions about the stability of the digital advertising market.

Business owners should view this not as an isolated technical error, but as a signal. The recent bug report and the subsequent replacement experiment prove that the interface between AI and the open web is still in a state of flux. For those scaling a business on the back of organic search and discovery, the priority must be the ownership of the audience. Relying on Google's "intended experience" is no longer a sustainable long-term growth strategy.

Further details on these shifts can be found via industry monitoring tools and community reports that track these flags in real-time.

FAQ

---

Sources: Searchenginejournal, Semasocial, Freeseoauditservices ·