Google's AI Search can now book your dinner

Planning group dinners usually means juggling three different apps, checking calendars, and manually matching party sizes with restaurant seating charts. Google wants its search bar to handle that entire headache for you.

Google rolled out agentic and personalized capabilities to AI Mode in Search, while expanding the core AI Mode feature in English to over 180 new countries and territories beyond its initial test markets in the U.S., U.K., and India.

Booking tables across platforms

The marquee update brings real-time task execution directly into Search. Google is starting with dining reservations, with plans to expand into local service appointments and event ticketing soon.

Users can toss out natural language requests packed with constraints like party size, specific dates, times, neighborhoods, and dietary restrictions. AI Mode queries multiple booking networks simultaneously, finds matching open tables, and delivers a curated list with direct links to finalize the reservation.

Under the hood, the system relies on the live web browsing capabilities of Project Mariner, Google Knowledge Graph, Google Maps, and direct integrations with platforms like OpenTable, Resy, Tock, Ticketmaster, StubHub, SeatGeek, and Booksy.

The catch is access. This agentic tool is currently confined to Google Labs for U.S. subscribers on Google AI Ultra. As TechCrunch pointed out, Ultra sits at the very top of Google's pricing ladder at a hefty $249.99 per month.

Memory from Maps and shareable threads

Alongside automated booking, AI Mode is getting personalized recommendations for Labs users in the U.S., starting with food. The engine looks at past conversations alongside locations tapped or searched inside Google Search and Google Maps to tailor answers.

If you ask for a quick lunch within an hour, and Google knows you lean toward outdoor seating and plant-based dishes, the recommendations adjust automatically. Users retain the option to manage or toggle this tracking in their Google Account settings.

Google also introduced a new link-sharing feature. Instead of copying and pasting chunks of AI answers, you can shoot a link to friends or family. Recipients jump straight into the conversation where you stopped and can continue querying the model to plan trips or events collaboratively.

Why it matters

For everyday users, AI Search is finally moving from summarizing text to executing multi-step tasks across the web. That said, locked behind a $250 monthly Ultra plan, this is currently a luxury demo for high-spending power users rather than a practical tool for the masses.

For the broader AI ecosystem, Google is flexing muscles that rivals like OpenAI cannot easily copy. Pure LLM companies can parse information, but Google owns the ubiquitous real-world infrastructure of Google Maps and established partner pipelines with ticketing and reservation giants.

We think stopping short of autonomous payment was a calculated, smart move. Linking directly to the checkout page protects Google from transaction liabilities while keeping partner platforms happy by preserving their user touchpoints. The real challenge begins when Google tries to scale these live multi-site agentic workflows to free users without collapsing under massive compute costs.