Google tests conversational AI search for YouTube Premium subscribers in US

Google is testing 'Ask YouTube,' an AI-powered conversational search interface for YouTube that generates text summaries alongside video results. The experiment is currently limited to YouTube Premium subscribers in the US who are 18 or older.

How Ask YouTube works

The feature adds an 'Ask YouTube' button to the search bar, with suggested prompts like "funny baby elephant playing clips" and "short history of the Apollo 11 moon landing." After submitting a query, YouTube displays a loading screen for several seconds before generating a page that includes:

• Text summaries with bulleted key points
• Timestamped video recommendations
• Organized galleries of related content
• YouTube Shorts in dedicated sections
• Follow-up question suggestions

Accuracy concerns remain

Testing by The Verge revealed factual errors in the AI-generated content. When asked about the Steam Controller, the system incorrectly claimed the discontinued original model had no joysticks, when it actually had one. YouTube correctly identified recently published content, including a Short from the same day as the test.

For some queries, such as "Apollo 11 conspiracy theories," the system reverted to standard YouTube search results rather than generating an AI summary page.

Expansion plans

According to YouTube, the company is "already working on" expanding the experiment beyond Premium subscribers. The feature follows Google's pattern of testing AI-powered search experiences, similar to AI Mode in Google Search and its integration into Gmail.

The interface resembles AI Mode's conversational approach, organizing information from multiple sources into a single, dynamically generated page rather than a traditional list of search results.

What this means

Google is applying its AI search strategy to YouTube's vast video library, attempting to make video content more accessible through natural language queries. The Premium-only rollout suggests Google is testing both the technology and potential monetization approaches. However, the presence of factual errors demonstrates that AI-generated summaries of video content still require human verification, particularly for technical or specialized topics.