On 6 August, Google expanded Ask Maps to more than 150 countries and territories. It is Gemini installed directly inside Google Maps: people no longer type "italian restaurant paris 11", they write "somewhere cosy for four at 7pm tonight, not too loud, where we can stay late". To build its answer, Google says it draws on more than 300 million places and reviews from more than 500 million contributors. Translated for a business owner: your listing, your reviews and your photos are no longer just read by a customer. They are the raw material for an answer written by a machine, on your behalf, and you will never get to proofread it.
What changed on 6 August
Ask Maps is not brand new. It launched on 12 March 2026 in the United States and India, on Android and iOS. What happened on 6 August is a different kind of event: it went global, alongside three additions that make Google's direction unmistakable.
| Feature | What it does | Where |
|---|---|---|
| Ask Maps | Answers in natural language, keeps the thread of a conversation, returns a custom map | 150+ countries, English only |
| Real time transit | Live bus, train, subway and ferry departures inside the answer, updated when delays hit | Global rollout |
| Personal Intelligence | Reads your Gmail (upcoming flight, dinner reservation) to tailor suggestions, opt in | Progressive |
| Agentic tasks | Food ordering (Square, Toast), hotel booking, event discovery | United States first |
Two honest caveats before going further, because they set the level of urgency. First: this wave is English only, and Google says it is working on other languages. Second: the headline features, ordering food or booking a hotel straight from the conversation, are still United States only. So if your customers do not search in English, they have not picked up this habit yet.
That is not a reason to close the tab. It is exactly the situation I described when Google started booking on behalf of your customers: when a feature actually lands, it lands with its conditions already written, and by then it is too late to prepare the ground. The useful window is now.
The question changed shape, and that is the real story
The important shift is not technical, it is linguistic. For twenty years a customer typed keywords, which meant guessing the machine's vocabulary. Today they describe their situation in their own words, and the machine has to do the adapting.
Google's own published examples say it plainly: "my phone is dying, where can I charge it without having to wait in a long line for coffee?", or "my friends are coming from Midtown East to meet me after work, any spots with a cosy aesthetic and a table for 4 at 7 tonight?". Look at what those sentences actually contain.
| What customers used to type | What they type now | What Google needs to know about you |
|---|---|---|
| italian restaurant paris 11 | an affordable italian for dinner for two around 8pm, quiet atmosphere | Price range, atmosphere, real opening hours |
| hairdresser paris 3 | a hairdresser open on Monday that takes walk ins | Exact hours, walk ins attribute |
| cafe wifi | a cafe where I can work for two hours with power sockets | Amenity attributes, reviews that mention working |
| bakery near me | a gluten free bakery open on Sunday morning | Dietary options, Sunday hours |
Only the right hand column concerns you. Every question of this kind carries an implicit filter, and that filter is almost always a field on your Business Profile that you either filled in or left empty. An empty field no longer costs you three places in a list. It removes you from the question.
That is the fundamental difference between a list and an answer. Eighth out of ten results is still a form of existence. In a sentence that names three places, you are either in it or you are not. I unpacked that selection mechanism in how a single question gets exploded into a dozen parallel searches before an answer is assembled. Same logic, now applied to the map.
What the answer is made of
Three raw materials, and only one of them is entirely yours.
Your listing attributes. These are the structured fields: outdoor seating, accessibility, payment methods, dietary options, pets, power sockets, reservations, walk ins. They are tedious to fill in, which is precisely why most listings leave half of them blank. Yet they are the only data a machine can filter on with certainty. It is the same principle as structured data on a website: describing facts in a format machines do not have to interpret.
The text of your reviews. When somebody asks for a cosy atmosphere, no checkbox says "cosy". That information comes from the vocabulary in your reviews. Which is why the star rating matters less every year while what your customers actually write matters more. A review saying "great" feeds nothing. A review saying "we stayed two hours talking and nobody rushed us" makes you eligible for a whole family of questions.
Your photos. They used to persuade a hesitating human. Now they also confirm an atmosphere, a type of venue, a layout. Google goes further still: in the new conversational contributions, a simple photo of your storefront can be enough to automatically detect the opening hours displayed in your window.
The AI summary block: Gemini describes you in your place
While everyone watches Ask Maps arrive, something quieter is already here. On mobile, just under the photos, some listings now show an automatically generated summary of the business. That summary was not written by the owner. Gemini produced it from what customers wrote in their reviews.
Take the measure of that. The first descriptive text a customer reads on your listing is no longer yours. It is an automated reading of your reputation, placed above everything you control. It does not appear everywhere either: Google reserves it for listings with enough written material for a summary to stand up. Which is itself a piece of information about what to do next.
You cannot write it, but you decide what it is made of. This block cannot be edited from your Business Profile interface, which is unsettling when you are used to controlling your own copy. The reasoning, though, is the same as for an AI citation: you do not write the answer, you write what the answer is built from. If your last twenty reviews talk about the welcome and the terrace, the summary will talk about the welcome and the terrace. If your most recent review is eighteen months old, the summary describes a business as it was eighteen months ago. A clear factual error can be reported to Google, but the real lever stays the raw material.
What this does not mean
I am not about to tell you everything has changed and you must act before Friday. That would be false, and you have read it twenty times elsewhere.
Ask Maps does not speak your customers' language yet in most markets. The integrated ordering and booking features are United States only. A large share of your customers will keep opening Maps, typing three words and scanning a list for a long time. Nothing below is a fire to put out this week.
What I am saying is narrower and easier to verify: the direction is set, it is consistent across all three surfaces at once, Google Search, Maps and assistants like ChatGPT, and the work it demands is identical in all three cases. Fill your listing to the last field, collect recent descriptive reviews, keep your hours honest. That work already paid off before. It simply became far more decisive. Which is also what makes it comfortable: nothing on the list below is a bet on a feature that might never ship in your market.
The four jobs, in order of return
1. Empty the list of unfilled attributes. Open your listing, go to the information section, and work through every attribute group to the end. Outdoor seating, wheelchair access, restrooms, payment methods, vegetarian and gluten free options, reservations, walk ins, pets, wifi, power sockets. Budget thirty minutes. It is by far the best effort to result ratio on this list, because every box you tick makes you eligible for an entire family of questions you were simply absent from.
2. Treat opening hours as live data. Half of all conversational questions contain a moment: tonight, Sunday morning, on Mondays, after 8pm. Wrong hours do not just cost you one customer, they teach Google that your listing is unreliable. Enter exceptional closures ahead of time, public holidays and staff leave included, and get into the habit of doing it the moment you decide them.
3. Change the question you ask your customers. Stop asking "would you mind leaving a review?". Ask "what did you enjoy?". The first phrasing produces five stars and zero words. The second produces text, and text is what makes you recommendable on a specific question. Then reply to those reviews, which lifts both your image and your local signals: that is what I cover in replying to Google reviews with AI in ten minutes a week, without losing your evenings to it.
4. Keep a website that gives something to cite. The listing carries the facts. The website carries everything else: your menu, your method, your group booking terms, your answers to the questions you get on the phone. That is the material assistants reach for when a question goes beyond the boxes of a form. If you are unsure where you stand there, the reasons a local business does not show up in ChatGPT make a good starting diagnosis, and the AI Overviews checklist for local businesses covers the Google side.
Once those four are done, take your baseline. Search Console's generative AI performance report gives you your impressions inside Google's AI answers: I explain what it shows and what it hides, because it is very easy to read the wrong way. And if your local SEO fundamentals are not in place, start there rather than with AI: the local SEO guide for shops and artisans comes before everything else.
The checklist
- Open your listing and fill in every attribute until none are left blank. Thirty minutes, once.
- Check your primary category and pick the most precise one that genuinely describes what you do, not the broadest.
- Enter your exceptional closures for the next three months, today.
- Change how you ask for reviews to an open question that invites a sentence, not a rating.
- Look at your AI summary block on mobile, if you have one. What it says tells you what Google has retained about you.
- Add five recent photos: storefront, room, products, team. Dated this year.
- Publish an FAQ page on your website with the real questions your customers ask, in their words.
Frequently asked questions
Is Ask Maps available in my country?
Probably, but likely not in your language yet. It launched on 12 March 2026 in the United States and India, then expanded on 6 August 2026 to more than 150 countries and territories, in English only, with Google saying it is working on other languages. The most advanced features, food ordering and hotel booking, are still United States first. So outside English speaking markets your customers have not built the habit yet, which gives you a few months to prepare your listing without pressure.
Can I edit the AI generated summary?
No, not directly. It is a Gemini summary built from your reviews, it is not written by you and cannot be edited from your management interface. The only real lever is changing what feeds it: recent, detailed reviews that talk about what you want surfaced. A clear factual error can be reported to Google, but do not treat that as your steering mechanism.
Do I still need a website if it all happens on the listing?
Yes, and the two do different jobs. The listing carries verifiable facts: address, hours, attributes, reviews. The website carries what goes beyond those facts and what assistants pull from when composing an answer: your menu, how you work, your terms, your answers to recurring questions. A listing without a website gives little to cite; a website without a current listing loses every proximity query. That is the whole point of a listing and a website properly wired together.
How many reviews do I need to count?
Google publishes no threshold, and be wary of anyone who gives you one. What is observable is that the automatic summary appears mainly on listings with enough written material. Two things matter more than the number: recency, reviews under six months old, and detail, reviews that tell a story. Ten recent, descriptive reviews beat a hundred old ones with no text.
What if I am not a restaurant?
The reasoning does not change, only the attributes do. A salon has walk in and accessibility fields, a tradesperson has a service area and payment methods, a practice has hours and contact options. The conversational question that concerns you already exists, it is simply phrased in your customers' vocabulary: listen to what they ask you on the phone, because those are the exact sentences people will type tomorrow.
Want to know what your listing actually says about you today, and what is missing for an AI to pick you? Get in touch: I look at your listing, your reviews and your website, and tell you what to fix first, in the order that pays best. That is the core of my work on AI search optimization, extended by AI for local businesses and, when the foundations are missing, by a website that finally gives machines something to work with. For a restaurant or a cafe, it all starts with the restaurant website page.