Google adds natural language controls to Discover feed for precise curation
A new chat interface allows users to describe exactly what they want to see in their feed, moving beyond simple algorithmic signals.
Google is introducing a chat interface for its Discover feed that allows users to customize their content stream using natural language prompts. This update shifts the power of curation from purely algorithmic signals to direct user intent, enabling a more nuanced way to manage the daily flow of information.
Instead of relying on binary "not interested" buttons or implicit signals from search history, users can now specify topics or links they want to see more or less of using their own words. For example, a user could request long-form content and deep dives regarding smart home technology while explicitly asking the system to skip product news and press releases. This natural language capability is part of a broader push toward personalization, which also includes the addition of customizable daily audio briefings by topic within the Google News app and an expanded "Preferred Sources" feature that allows users to prioritize specific publishers.
The shift from implicit to explicit control
Historically, the Google Discover feed has operated as a largely "black box" experience. The algorithm determines content based on a user's search history, visited pages, and YouTube activity. While users could block specific publishers or mark individual stories as irrelevant, these tools were often too blunt for fine-tuning. A user interested in a specific product model, for instance, might find themselves overwhelmed by general news about a brand because the algorithm could not distinguish between a niche interest and a broad category.
Why intent-based curation matters
This transition toward "intent-based" curation represents a significant change in how AI-driven discovery functions. By giving users agency over the process, Google is attempting to reduce the noise and friction often associated with algorithmic feeds. The ability to provide specific instructions—such as requesting tips for a specific device while ignoring general brand news—allows for a higher signal-to-noise ratio. For the industry, this move suggests that even the most powerful recommendation engines are reaching a limit where human-defined boundaries are necessary to maintain user satisfaction and engagement.
What to watch
As these features roll out, the primary question remains how effectively the AI will interpret complex or contradictory natural language requests over time. While the system is designed to remember these preferences for future delivery, the precision of this "memory" will determine if the feature becomes a primary tool for users or a novelty. Additionally, the impact of the "Preferred Sources" feature on publisher traffic remains to be seen, as it may further consolidate the influence of established outlets that users explicitly choose to prioritize.