For eight years, a customer could type a question onto your Google listing and you could type the answer back. That box is gone. Google shut down the Business Profile Q&A API on November 3, 2025, and on March 12, 2026 it replaced the feature with Ask Maps — a Gemini-powered answer that gets written on the spot, from your attributes, your services, your hours, your photos, your reviews, and your website. The question still gets asked. You just don't get to answer it anymore. Something else answers it, using whatever you left lying around on your profile.

For a shop, salon, studio, or clinic, that is a bigger change than it sounds. The questions people ask before they walk in — do you take walk-ins, is there parking, do you carry that brand, do you do curly hair, ¿hablan español? — used to be answerable in one sentence by the owner. Now they are answerable only in advance, and only in the fields the model reads.

The math

The feature died of neglect, and the numbers explain why Google felt free to remove it. A ThriveHive analysis of questions across more than 43,000 Google My Business profiles found that 91% of questions went unanswered by the business owner, and that 82.6% of the questions that did get an answer were answered by a Google Local Guide — a stranger, not the business. A separate GatherUp study found roughly a quarter of listings had been asked at least one question at all.

So for most local businesses the practical situation was already this: someone else was answering for you. The 2026 change didn't create that problem. It industrialized it, and made the answer harder to argue with, because it now arrives in Google's own voice instead of as "Answered by a Local Guide, 3 years ago."

What makes this urgent rather than merely interesting is where discovery is heading. BrightLocal's Local Consumer Review Survey 2026, run on a representative panel of 1,002 US adults, found that the share of consumers using AI tools — ChatGPT, Google's AI Mode, Gemini and peers — to find local businesses jumped from 6% in 2025 to 45% in 2026, making AI the third-largest discovery channel, ahead of Yelp and Tripadvisor. The same survey found 97% of consumers read reviews before choosing a local business, and 41% now say they always do.

The old skill was answering questions well. The new skill is making sure the answer already exists, in a field a model can read, before the question is asked.

Why Florida indies specifically

Three things make this land harder on independent retail here than in most of the country.

Language. Most Miami-Dade households speak a language other than English at home. Your customers ask questions in Spanish, Portuguese, Creole and Russian, and the model answers in the language of the question — drawing on text you may only have written in English. A salon in Doral with twenty Spanish reviews describing alisado and balayage is legible to that model. The same salon with twenty five-star reviews that say "great!" is not.

Foot traffic that decides in the street. On Miracle Mile in Coral Gables, in Aventura, on a Gainesville block near campus, the question gets asked from the sidewalk, thirty feet from your door, on a phone. There is no second visit to your website to clear up an ambiguity. If the answer says "appointments recommended" and you actually take walk-ins all Tuesday, you lost a customer who was already standing outside.

Seasonality that outruns your profile. Snowbird hours, summer hours, holiday closures, hurricane closures. A model reading a stale hours field will confidently tell someone you're open. Under the old system a customer could ask and you could correct it. That correction path no longer exists.

The playbook

Step one: harvest the ten questions from the counter, not from a keyword tool

For one week, have whoever works the front write down every question a customer asks out loud, on paper, by the register. Do not filter. You will end up with fifteen to twenty, and ten of them will repeat daily: parking, walk-ins, price range, do you take Apple Pay, do you have it in stock, how long does it take, do you do kids, is there someone who speaks Spanish, can I bring my dog, do you validate.

This list is the specification. Everything below is just placing those ten answers where the model can find them. Keyword research will not produce this list, because these are not searches — they're the friction that stops a purchase.

Step two: put every answer in a structured field, not in a sentence

This is the step people get wrong. Writing "we take walk-ins and have free parking behind the building!" in your business description is nearly useless — it's one unweighted blob of prose. Google gives you dedicated, machine-readable fields, and they carry far more weight:

  • Attributes — accepts walk-ins, wheelchair accessible entrance, free parking lot, restroom, LGBTQ+ friendly, women-owned, gender-neutral restroom, appointment required.
  • Payments — credit cards, debit, mobile NFC, cash-only. "Do you take Apple Pay" is a top-ten question in every shop we audit and takes eight seconds to answer permanently.
  • Hours, including special hours — the separate field for holidays and one-off closures. Set Thanksgiving, Christmas, July 4th and your summer schedule now, in one sitting, for the next twelve months.
  • Accessibility and amenities — the fields that answer the questions people are least likely to ask you directly and most likely to ask a phone.
  • From the business — identity attributes that increasingly show up in AI summaries.

Every attribute you leave blank is a question the model has to guess at, or answer from a three-year-old review.

Step three: name your services the way customers say them

Your services and products list is now one of the model's primary sources, and most independents fill it in with internal shorthand. "Color — Level 2" means nothing to a model matching a question about balayage. "Arrangement, medium" doesn't match "same-day delivery for a funeral."

Write each service the way the customer says it out loud, then add one sentence of description that includes the qualifier they'd use: the hair type, the occasion, the age group, the brand, the turnaround, the language. A florist in Brickell listing "Same-day arrangements — orders placed before our posted cutoff, delivered across Brickell, downtown and Coral Gables" is answering four questions at once, in a field built to be read.

Products deserve the same treatment. If you carry a brand people search by name, the brand belongs in the listing, not just on the shelf.

Step four: make your reviews specific enough to be quoted

The model quotes reviews. That's the mechanism — it reads what customers wrote and paraphrases it back to the next customer. Which means a wall of "Great service!" five-stars gives it nothing to work with, while a handful of detailed reviews can answer questions you never filled in anywhere.

So change what you ask for. Instead of "please leave us a review," ask for the detail: "If you have a second, would you mention what you came in for?" That single sentence, said at checkout, is the difference between a rating and a source. And when a review comes in Spanish, reply in Spanish — the reply is indexed too, and it is the cheapest possible signal that someone at your counter speaks the language.

One caution: never write or solicit a review that describes something that didn't happen. Beyond being against Google's policy, it now feeds a machine that will repeat it to a customer who will arrive expecting it.

Step five: ask the question yourself, once a month

This is the only feedback loop you have left, and it takes four minutes. Open Maps on a phone that isn't signed into your business account and ask the three questions that matter most to your revenue, phrased the way a customer would: "does [your shop] take walk-ins," "is there parking at [your shop]," "does [your shop] do [your highest-margin service]." Read what comes back.

You are not looking for a good review. You are looking for a wrong fact — the outdated hours, the service you stopped offering, the "appointment required" that isn't true anymore. When you find one, you don't argue with the answer; you go fix the field it came from. Put it on the calendar for the first Monday of every month.

What good looks like

A salon in Coral Gables. Ten counter questions collected in a week. Attributes set: accepts walk-ins, free street parking nearby, wheelchair accessible entrance, mobile payments. Services rewritten from "Color 1 / Color 2 / Color 3" to "Balayage," "Curly cut — dry cutting for 2B–4C hair," "Blowout — 45 minutes," each with a one-line description. Checkout script changed to ask customers to mention what they came in for. Six weeks later the question "does anyone in Coral Gables do curly cuts" has something true to match against.

A boutique in Aventura. Special hours entered for the next twelve months in one sitting, including the two weeks it closes in August. Payments and parking attributes filled. The three brands customers ask for by name added to the products list. The website gets a plain FAQ block — one question per heading, answered in one or two sentences, naming the shop and the neighborhood, because the linked website is one of the sources the model reads.

The monthly check. A note in the owner's calendar: first Monday, four minutes, three questions asked on a personal phone. In March it caught hours that still showed the holiday schedule. That is the entire maintenance cost of the system.

The part nobody will do for you

Every one of these steps is free. None of them requires a developer, an agency, or a subscription. What they require is an hour of somebody's attention and the willingness to treat a set of dropdown fields as customer service, which is exactly what they now are.

The businesses that will lose ground over the next year aren't the ones with bad answers. They're the ones with blank fields, three-year-old hours, services named in internal shorthand, and forty reviews that all say "great." A model reading that profile will still produce an answer — it just won't be yours, and you won't get to correct it.

If you want to know what your profile is currently telling that model, we'll show you. Boost My Spot runs a free five-page audit, delivered in 48 hours, with no sales call — your attributes, services, hours, review language, and the gaps a machine is currently filling in with guesses. Salon, clinic, boutique, studio, something we haven't listed? The problem is the same one. Request it at boostmyspot.com/audit.

Sources