A year ago, if you wanted a good physio near the office, you searched Google, scanned the map, and read a few reviews. Plenty of people still do. But a growing number now just ask an assistant: "Who's a good physio near Liverpool Street that's open on Saturdays?"
In BrightLocal's 2026 Local Consumer Review Survey, 45% of consumers said they'd used ChatGPT or another generative AI tool for local business recommendations. The year before, it was 6%. That's not a trend you can wait out.
So what decides whether your name comes up in that answer?
A note of honesty first
Nobody outside the companies that build these assistants knows exactly how they rank businesses, and the systems change often. Anyone who promises a guaranteed spot in AI answers is guessing or selling. What we can do is look at what these tools say they use, what they cite, and what they consistently pick up, and work from there.
What assistants seem to lean on
Reviews, and what's written in them
When an assistant explains a recommendation, it usually quotes or paraphrases reviews. "Customers mention short wait times and a friendly front desk." That means the words in your reviews matter, not just the star average.
A review that says "Great service!" gives an assistant nothing to work with. A review that says "They fitted me in on a Saturday morning and explained the exercises properly" answers exactly the kind of question people ask.
You can't write reviews for customers, and you shouldn't steer what they say. But asking them a specific question in your request ("What did we help you with today?") tends to produce more detailed reviews naturally.
Recency
Assistants often favour current information. A business with steady recent reviews looks open and active. Fifty reviews from three years ago and nothing since can look like a place that's closed or slipped. Building a weekly habit of asking, as we describe in how to get more Google reviews within the rules, helps here.
Consistent facts across the web
Your name, address, phone number, hours and categories should match everywhere: your website, Google Business Profile, Apple Maps, Bing Places, Yelp, industry directories. When sources disagree, an assistant has to choose, and it might choose to leave you out rather than risk being wrong.
Spend an afternoon checking. It's dull and it matters.
Your replies
Replies show up alongside reviews on most platforms and can be read like any other text. A business that answers questions and concerns in public gives an assistant more context about how it operates. Half of consumers in the BrightLocal survey said templated replies put them off, and generic replies also give assistants nothing useful. Our guide to replying to negative reviews shows how to write replies that sound like a person.
A clear website
Assistants that can browse often visit your site. Make the basics impossible to miss: what you do, where, for whom, your hours and how to book. Plain text beats clever design here. Structured data for your business type, address and opening hours helps machines read it correctly.
Questions people actually ask
It helps to think in the questions, not the keywords.
- "Which dentist near me is good with nervous patients?"
- "Best bakery in Clapham for a birthday cake order?"
- "A plumber who can come today in Manchester?"
For each question you'd like to be the answer to, ask: is there evidence on the web, in my own words and my customers' words, that I'm good at this? If not, that's your gap.
Things that don't help
- Stuffing keywords into replies. "Thanks for visiting the best dentist in Leeds" reads badly to people and does nothing useful.
- Fake or incentivised reviews. Beyond being against the rules (see what the FTC fake review rule bans), they tend to be generic and get filtered.
- Chasing every new AI tool separately. The fundamentals above work across all of them.
Trust signals are rising in value
As assistants summarise reviews instead of showing them all, the reliability of those reviews becomes more important. If the source material is full of fakes, the summary is too. That's one reason review verification is getting attention. We explain the honest version of it in what blockchain review verification actually proves.
A one-hour checklist
- Search your business name in two or three assistants and note what they say, including anything wrong.
- Fix inconsistent hours, addresses and categories across your main listings.
- Reply to your last ten reviews that don't have a reply.
- Change your review request to ask one specific question.
- Add a plain "what we do, for whom, where" paragraph to the top of your homepage.
Then check again in a month. It moves slowly, but it moves.
