Searching for a restaurant on a map has meant, for twenty years, typing three words and choosing among ten red pins. On November 17, 2026, Google rolled out Ask Maps in Spain and in more than thirty European countries, a conversational layer over its maps service that changes the way people ask: instead of a search, a sentence with the whole plan inside it.
The update arrives with Gemini behind it and with the ambition of being the biggest change to the service in more than a decade: the Ask Maps demo in Madrid showed the full itinerary drawn on the map, with numbered stops and the plan exported to a document. The user describes what they want to do, and the system returns an itinerary with numbered stops, opening hours, predicted traffic and the option to add a new stop midway.
The rollout is gradual and covers both Android and iOS, so not every phone will see the feature on the same day. The figures the company puts on the table to explain the scale are 300 million places and reviews from 500 million people who have contributed to the map over the years, a corpus that no model can fabricate on its own and that turns an answer into a defensible recommendation.
What can you ask for and what does it return?
That database is what makes it possible to move from answering where a place is to proposing an entire plan. The example Google chose to show it off works as a statement of intent.
The demo the company used as an example is a three-day getaway along the coast with stops for hiking and eating by the sea. The system returns the day-by-day schedule that Google describes with tips and numbered markers on the map, so the plan can be followed on the phone and exported to a document to share with fellow travelers.
The second feature is not about planning but about querying what has already been saved. Ask Maps lets you ask about places on your own lists, such as which restaurant has a table free for five people in the next thirty minutes, and it answers by cross-referencing that list with the availability information the service already collects.
The third is live information during navigation. The assistant warns about traffic, opening hours and the expected crowds at the next stop, and accepts changes on the fly without redoing the whole plan. Anyone who has improvised a route with three apps open knows the time saved.
The fourth element is export. The itinerary leaves the map as an editable document, which sounds minor and solves one of the classic problems with lists of places: the plan is shared with someone who does not have the app installed and is printed on paper when coverage fails.
There is a fifth piece the company mentions in passing and that explains the launch timing. The new features are designed for the year-end shopping and travel season, when the volume of queries about places and opening hours multiplies, and that is when a planner stakes its reputation.
The map as a conversational interface
What matters about Ask Maps is not that it understands long questions, but that it abandons the logic of the single query. A map has always worked as a search engine for specific places, and the assistant treats it as a planner of sequences where order, distance between stops and the time of arrival at each one matter.
That shift has consequences for the kind of error the system makes. A search that fails returns an irrelevant list and the user notices immediately, whereas an itinerary that fails returns a plausible plan with a stop that is closed on Mondays or with two hours of traffic jams between the second and third visits.
For Google, the move has a clear product logic: if the assistant plans the trip, advertising and bookings are placed inside the conversation instead of at the end of a list of results. The map stops being a catalog and becomes the interface where decisions are made.
The company has not detailed how it verifies opening hours or how often it updates availability, two data points that decide whether the plan can be executed or is only good for inspiration. Nor has it explained what happens when the assistant gets it wrong and the user has already booked the table at the wrong place.
What does the figure of 300 million places reveal?
The number matters less for its size than for what it implies about dependency. An assistant that plans trips needs up-to-date data on millions of places, and that database is the competitive advantage no language model can replicate on its own, however good it is at writing itineraries.
The other half of the equation is reviews. Five hundred million people have left opinions on the map over the years, and that corpus is what makes it possible to answer whether a place is good to go to with children or whether the terrace is worth it in winter. Without that material, the assistant could only sort by distance and open the door.
The obvious risk is the quality of what it recommends. Reviews have been a breeding ground for commercial abuse for years, and a system that summarizes opinions can amplify a bias if it does not distinguish between a hundred genuine ratings and a hundred bought in a single week.
The third consequence is about the market. With an assistant that proposes itineraries, traffic to travel websites changes hands, and local businesses that depended on appearing in a search now depend on being included in the plan. It is the same tension the web already lives with over AI-generated summaries, and the same question about who pays for the content the assistant summarizes.
Travel guide publishers noticed it before anyone else. A page with twenty recommended restaurants loses its point when the system answers with the three that fit the plan, and the traffic that used to arrive through the long tail concentrates in the assistant's answer.
The assistant that replaces the list of places
The arrival of Ask Maps in Spain closes a year in which Google has moved its search engine and its maps into the same territory. The Gemini 4 Hydrogen model already changed the way queries are interpreted in search, and now the same approach lands in the place where people decide where to eat and how to get there, with the same logic of long answers that the search engine debuted in November.
For the user, the useful question is whether the itinerary survives contact with reality: a three-day plan with closed stops or outdated hours is abandoned at the first detour, and then the assistant will have saved twenty minutes of searching only to cost an afternoon of corrections.
For the local business, the question is different. Appearing on the map will no longer be enough if the assistant decides what fits the plan, and anyone who wants to be in that answer will have to look after the data in their listing with the same care they give their website. The map has become an intermediary that has opinions, sorts and discards.