Google is bolting a chatbot onto its Discover feed. In the coming days, users of the Google app will find a new option in Discover's three-dot menu that opens a conversational interface for describing, in plain language, what they want the feed to surface. The chatbot confirms the request, lists the content types it plans to prioritize, and lets the user add corrections before hitting a "Refresh your feed" button to apply the changes. The system remembers those preferences for future visits.
The move turns Discover, which until now inferred interests from a user's activity across Google Search and other Google apps, into something a reader can steer directly. Instead of clicking thumbs-up and thumbs-down on individual cards and hoping the model learns, users tell it what they want. It is a small interaction change with a larger implication: Google is treating natural-language prompts as the default control surface for its consumer products.
Discover is one of the most-used pieces of the Google app on Android, feeding hundreds of millions of users a personalized card stream every day. Whatever changes the ranking there changes what a significant slice of the mobile web reads. Handing users a chatbot to reshape that ranking is a bigger lever than the feature framing suggests.
Key facts
- 01Google is adding an AI chatbot inside Discover that lets users describe their content preferences in natural language.
- 02The feature rolls out in the Google app in the coming days, accessed via the three-dot menu on the Discover feed.
- 03Users can hit 'Refresh your feed' after the chatbot confirms preferences; the system remembers choices for future visits.
- 04Google News audio briefings on Android now let users pick which topics to hear about.
- 05Preferred Sources gets an embeddable button publishers can place on their sites to let readers subscribe without leaving the page.
Google is not alone in this pattern. YouTube, Instagram, Bluesky, and X have all shipped AI-driven ways to reshape a feed in the past year, each betting that opaque recommendation systems become more defensible when users can talk to them. The wager is that a feed a user can argue with feels less like a black box, even when the underlying ranking model is unchanged.
Alongside the Discover update, Google announced two other tweaks. Daily audio briefings in the Google News app on Android are now customizable by topic, so users can decide which subjects the briefing covers rather than accepting a default mix. The change moves audio briefings closer to a self-assembled podcast than a general news readout.
Google also updated Preferred Sources, the feature that lets readers pin favored outlets to the top of Search's top-stories module, AI Overviews, and AI Mode. Publishers can now embed an interactive Preferred Sources button on their own sites, letting a reader add the outlet to their list without leaving the page. That closes a friction gap: previously, adding a source required navigating back into Search settings.
The Preferred Sources button lands in the middle of a running fight between Google and publishers over what AI-generated summaries are doing to referral traffic. AI Overviews and AI Mode answer many queries on the results page, and publishers have been vocal that clickthroughs to their sites are down as a result. AI Chat Daily covered Google's earlier Preferred Sources rollout last week, when the feature was pitched as a way to give publishers a fighting chance at visibility in an AI-summarized results page. The embeddable button extends that lifeline by letting outlets recruit their existing audiences into the feature directly.
Whether that meaningfully offsets AI Overviews' impact on publisher traffic is a separate question. Preferred Sources influences which outlets appear in the top-stories module and inside AI-generated answers, but it does not change the underlying dynamic where the answer sits on Google's page rather than the publisher's. The button is a distribution assist, not a business-model fix.
The Discover chatbot raises its own second-order questions. A feed users can describe in plain language is also a feed users can accidentally narrow into an echo chamber, and the system's willingness to "remember" preferences means those narrowings persist. Google's video demo shows the chatbot confirming choices back to the user, which at least creates a chance to catch overly restrictive prompts, but the tuning still puts the ranking decision closer to user whim than before.
For Google, wiring a chatbot into Discover is a low-risk way to keep pushing conversational interfaces into products that already have hundreds of millions of daily users. Discover is not a moonshot; it is one of the most-touched surfaces the company owns. Getting users comfortable typing preferences into a Google-branded chat box there is training data, habit-formation, and a foothold for the broader push into AI Mode across Search. The Discover update looks small on its own, but read alongside Preferred Sources and audio-briefing personalization, it is the same play in three different rooms: convert every ranking decision into a conversation, and make Google the one you have it with.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
