Journal / AI Marketing

How AI Assistants Decide Which Local Businesses to Recommend

Illustration of an AI assistant choosing among local business storefronts on a map

Someone in Richmond opens ChatGPT and types “best dentist in Richmond BC.” The assistant thinks for a few seconds and names three or four clinics. That person will probably call one of them — and the twenty other dentists in the city will never know the conversation happened.

We wrote a full guide to Generative Engine Optimization covering how brands earn AI citations in general. This article is the local companion: how assistants handle “near me” questions specifically, and what a local business can do about it.

What actually happens when you ask an AI for a local recommendation

Modern assistants don’t answer local questions from memory alone. For a query like “best dentist in Richmond,” ChatGPT, Claude, and Perplexity typically do two things:

  1. Retrieve — run a live web search and pull in pages that match the service and the location: business websites, review platforms, directories, articles.
  2. Synthesize — compress those pages into a short answer, naming the businesses that appear most clearly and consistently across the retrieved sources.

There is no local pack, no map widget, no page two. The synthesis step is brutal: a handful of names make the answer, and everyone else is invisible.

How many people actually ask? In a July 2026 BrightLocal survey of 1,227 US consumers, reported by eMarketer, 31% said they now use AI monthly for local recommendations, up from 6% a year earlier. But 75% of those AI users checked Google or Maps before contacting the business, and only 2% relied on AI alone. It is US data, and Canada may lag, but the pattern matters: AI is becoming the first step, and your Google presence is still the confirmation step.

The sources AI leans on for local answers

In our own testing across assistants, local answers tend to draw on a recognizable set of sources:

Notice what’s missing: your ad budget. AI answers are, for now, an organic channel — the assistant recommends whoever the evidence supports. Accuracy varies too: a 2026 study by the vendor SOCi, covering about 350,000 locations of multi-location brands, found Gemini’s business information was fully accurate, which Search Engine Land attributed to its grounding in Google Maps, while ChatGPT and Perplexity were right about 68% of the time (Search Engine Land). Assistants that are not reading a maps database are relying on whatever the web says about you, which is why consistency matters so much.

The cost of being invisible

Here is what makes AI invisibility different from a low Google ranking: there is no feedback signal. With traditional search you can see impressions and clicks slipping. With AI recommendations, the customer asks, gets an answer that doesn’t include you, and books a competitor. No analytics event fires. You lose the customer before you ever knew they were looking.

That’s why we treat AI visibility as a standing audit item, not a one-time project.

The local AI-visibility checklist

1. Say what you do and where, in plain sentences

Your homepage and service pages should contain liftable statements: “We are a family dental clinic in Richmond, BC, offering cleanings, implants, and emergency care.” An AI can quote that. It cannot quote a hero banner that says “Smile brighter.”

2. Keep your name, address, and phone identical everywhere

Assistants cross-reference sources to confirm an entity. If your website, Google Business Profile, Yelp page, and directory listings disagree on your suite number or phone format, you dilute the very signal that gets you named. Pick one canonical version and enforce it. Our citation map guide lists where those listings live.

3. Publish FAQ pages built from real customer questions

The questions people ask assistants are the questions they used to ask your front desk: “Do you take new patients?” “Do you offer direct billing?” Answer them on your site in self-contained paragraphs — that’s exactly the shape of content AI lifts into answers.

4. Earn reviews with substance

A hundred five-star reviews that say “Great!” teach an AI almost nothing. Reviews that mention the specific service, the neighbourhood, and the outcome give the model quotable evidence.

Be careful how you get there, though. Since February 2026, Google’s review policy says businesses should not pressure customers to review while they are on the premises, or request that specific content be included in a review (Google policy). So don’t script reviews. Ask at the right moment instead, right after a specific job or visit, and make it easy. A review requested while the experience is fresh naturally has more to say than one requested months later. (We’ve corrected this section; an earlier version suggested asking customers to mention what they came in for, which the updated policy rules out.)

5. Get mentioned on other local sites

Sponsor a community event, join the local business association, pitch a local publication. Every independent mention corroborates that you exist, where you operate, and what you’re known for — and corroboration is what models trust. Our local partnerships guide covers how to pick and structure those.

6. Let the AI crawlers in

None of this helps if the assistant can’t read your site. Check that your robots.txt allows the search crawlers, especially OAI-SearchBot, which OpenAI says sites must allow to appear in ChatGPT search answers. Our GEO guide lists the crawler names and what each one controls.

Audit your AI visibility monthly

Once a month, ask ChatGPT, Claude, Gemini, and Perplexity your money questions — the queries a ready-to-buy customer would type. Log three things: whether you’re named, how you’re described, and who is named instead. The gaps tell you exactly which fact to publish, which listing to fix, or which review pattern to encourage next.

This is the new rank tracking, and most of your competitors aren’t doing it yet.

If you’d like us to run an AI-visibility audit for your business and build the fix list, that is part of our local SEO and AI search service, or email Charlene.

Frequently Asked Questions

How does ChatGPT pick which local businesses to recommend?

When asked a local question, assistants like ChatGPT and Perplexity typically search the live web, retrieve pages that clearly match the service and location, and synthesize an answer from them. Businesses whose websites, reviews, and directory listings state plainly what they do and where they do it are the ones that get named.

Why does my business not show up in AI answers?

The most common causes are a website that never states your services and city in plain sentences, inconsistent name, address, and phone details across the web, thin or generic reviews, and few third-party mentions. AI systems cannot recommend an entity they cannot clearly identify and locate.

How can I check my own AI visibility?

Once a month, ask ChatGPT, Claude, Gemini, and Perplexity the exact questions your customers would ask — "best [your service] in [your city]" — and log whether you appear, how you are described, and which competitors are named instead. Treat it like rank tracking and use the gaps to decide what to publish next.

Do people actually use AI to find local businesses?

More every year. In a July 2026 BrightLocal survey of 1,227 US consumers, reported by eMarketer, 31% said they use AI monthly for local recommendations, up from 6% a year earlier. Most still double-check: 75% of those AI users looked at Google or Maps before contacting the business, and only 2% relied on AI alone. That makes your Google presence the confirmation step even when AI makes the first suggestion.

Keep reading

Get in touch

Let's talk about your brand