How to get your business recommended by ChatGPT

By Akshay · August 1, 2026 · 9 minute read · For single-location owners

Someone in your town opens ChatGPT and types "best hair salon near me." They get two or three names and a sentence about each. They call one. You never find out it happened.

That is the entire problem in three sentences, and it is why this question has started coming up on sales calls. So here is the honest version of the answer, including the parts that are inconvenient.

First, the uncomfortable number

A 2026 analysis of more than 350,000 business locations found that ChatGPT named just 1.2% of them when asked for a local recommendation. The same businesses appeared in Google's local map pack 35.9% of the time. Roughly a thirty-fold gap.

Two things follow from that. Being recommended is rare, so the bar is lower than it looks once you actually try. And your Google ranking does not carry over, so you cannot assume you are fine because you show up on Maps.

What ChatGPT is actually doing

It is not consulting a ranking table. When someone asks for a local recommendation, the model is assembling an answer from what it can find about businesses in your area: structured data from Google Business Profiles, the text of reviews, and whatever the wider web says. Then it names the two or three it is most confident about.

Confidence is the word that matters. A model does not want to recommend a business it cannot describe. If your profile is half empty, your reviews are thin, and your details disagree between three directories, you are the risky answer. It picks someone else.

Six things you can actually change

1. Finish your Google Business Profile

Not "claim it." Finish it. Primary category correct, secondary categories added, every service listed with a description, attributes ticked, hours including holiday hours, photos uploaded recently, and the questions section seeded and answered. Most profiles we inherit are around half complete, which is the cheapest advantage available to you.

2. Get reviews consistently, not in bursts

Recency carries real weight. Forty reviews spread evenly across two years reads as an active business. Forty reviews from one week in 2023 reads as a campaign that ended. A steady trickle beats a spike every time.

3. Get reviews that say something

This is the one most owners miss. A five-star review with no text gives an assistant nothing to quote. A review saying "took my BMW in for a coolant leak, quoted $340, charged $340, done same day" gives it a sentence it can use for three different questions. When you ask for reviews, ask people what they came in for.

4. Reply to everything

Your response rate is visible and it signals an active business. It also doubles the text about you, in your own words, which is your only chance to state things your customers did not think to mention.

5. Make your details agree everywhere

Same name, same address format, same phone number, same hours, on your site, your profile, and every directory you appear in. Business information is only about two thirds accurate across the major assistants. Every contradiction lowers confidence.

6. Answer the questions people actually ask

Assistants answer questions, so they favour businesses whose web presence answers questions. Not a services page. Actual answers to what people in your area ask: what it costs, how long it takes, whether you handle their specific situation.

What does not work

Buying a guarantee. Nobody can guarantee placement, answers shift with phrasing, and anyone promising otherwise is telling you something about themselves.

Stuffing "AI optimised" onto your homepage. There is no keyword to rank for here. The model is reading about your business, not your marketing copy.

Gating reviews so only happy customers reach Google. It violates Google's policies, sits badly with the FTC's rules on review suppression, and produces a suspiciously uniform review profile that helps nobody.

The ten-second version

Open ChatGPT. Ask for the best business in your category and town. See if you are named, and see who is. That second list is more useful than your own result, because it tells you exactly who the model currently considers the safe answer in your market.

Then go and be more documented than they are. That is genuinely most of it.

Common questions