Google I/O puts Gemini into search and ships Veo 3

At its annual I/O developer conference, Google announced that AI Mode — a conversational, agentic search interface distinct from the AI Overviews summaries already embedded in results — would roll out to all Search users in the United States, having previously been available only to Labs testers. The company said AI Overviews, the more limited AI-generated summary feature, had reached 1.5 billion monthly users across 200 countries and territories since the prior year’s conference. AI Mode added a “Deep Search” option for more thorough, multi-step research responses and agentic capabilities for tasks such as booking restaurant reservations, alongside a “Search Live” feature planned for that summer that would let users hold real-time conversations using their phone’s camera.

The event also introduced Veo 3, an updated video-generation model that, unlike its predecessor and most competing video models, generated synchronised audio — dialogue, sound effects and ambient noise — alongside the video rather than producing silent clips. Google paired it with editing tools carried over from Veo 2, including camera controls and object manipulation.

Google also detailed updates to its underlying language models: Gemini 2.5 Pro, which it said led its performance benchmarks and gained a “Deep Think” mode for more extensive reasoning, and a faster Gemini 2.5 Flash with improved coding output. The company said more than seven million developers were building with Gemini. Taken together, the announcements marked Google folding generative AI more deeply into Search itself, the product on which most of its advertising revenue depends, rather than treating chat-based AI as a separate destination from search.