Google Tests "Dive Deeper" AI Button in Discover Feed to Surface More Publisher Sources

Google has launched a live experiment introducing a "Dive deeper" interactive button inside its Discover feed, pairing AI-generated topic summaries with curated rows of source cards and outbound links to original reporting. The test was publicly disclosed by Robby Stein, Vice President of Product for Google Search, via a post on X.

According to the source, the experiment aims to help mobile users explore complex topics and discover diverse perspectives across the web without having to repeatedly execute manual search queries. Currently, the test focuses primarily on video content.

How the Feature Works

When a user encounters a qualifying video in their Discover feed, they are shown a short AI-generated summary of the topic. A disclaimer accompanies the text noting that the summary has been generated with artificial intelligence and is subject to potential inaccuracies. Below that summary sits the new "Dive deeper" button.

Tapping the button opens a dedicated overview interface featuring a more robust, expanded summary accompanied by a curated row of prominent source cards and direct links to external reporting. In examples provided by Google, the initial test case used a local news weather video concerning a Northern California heat wave. The expanded layout organises information under an "Explore more stories" banner, with sub-topic headlines, concise descriptive text, and outbound links to various original publishers and community platforms.

Google has not yet disclosed specific geographical regions, language parameters, or user rollout percentages associated with this initial testing phase.

Part of a Broader AI Rollout in Discover

The "Dive deeper" button is part of a wider deployment of generative AI features within Discover. In October 2025, Google deployed AI-powered previews on trending topics within the Discover feed, initially focusing on regional markets including the United States, South Korea, and India.

Separately, on August 31, the company rolled out a dedicated generative AI performance report within Google Search Console specifically for Discover traffic. The report aggregates data categorised by page, country, and date, utilising impressions as its primary metric. However, the report does not yet break down performance metrics by specific individual features — meaning publishers cannot yet distinguish between standard AI previews and the newly tested "Dive deeper" interface in their analytics.

Google has also clarified that experimental features running exclusively through Search Labs are typically excluded from these standard performance metrics, though it remains undetermined whether the current "Dive deeper" test falls under Search Labs or general production experiments.

Implications for Publishers

The architectural design of the feature — with an explicit row of source cards linking to original journalism — is presented as a response to ongoing industry concerns about AI-driven search features that traditionally keep users within search engine result pages. By highlighting multiple source domains, the feature is designed to act as a gateway for users to consume original reporting from a diverse spectrum of outlets.

Nevertheless, the absence of granular Search Console reporting currently makes it difficult for digital marketers to measure the precise traffic impact of the "Dive deeper" button. Publishers will likely need to rely on direct referral log analysis to gauge how effectively the new interface translates into actual user visits.

Robby Stein noted that engineering and product teams intend to test multiple design variations over the coming weeks to determine which user interfaces yield the most intuitive and satisfying experience. While the current phase is confined to video content, expansion to standard textual articles and other content formats is anticipated by industry analysts, though Google has not provided a definitive timeline.