Skip to main content
Back to blog

By On Point Tech Solutions

People Ask ChatGPT For Recommendations. Here's What It Reads

On Point Tech blog cover about what ChatGPT reads when recommending a business, with Michael in a dark suit on a light purple background

Where should I take my car. Who do I call about a leak. Which place nearby does a decent haircut. People are typing those into an AI assistant now instead of into a search bar, and they are getting back a real answer with real business names in it.

The obvious question is how your name gets into that answer. The useful question is a level down from that: where is the answer coming from in the first place?

Because none of these systems knows anything about your business on its own. It has no private file on you. Everything it says about you, it got from something public. Once you see the list of what that actually is, the work stops being mysterious and starts being a checklist.

It reads what everyone else reads

Strip away the technology and the sources are the same places a nosy customer would look. That is genuinely the whole secret.

Your Google Business Profile. The most structured thing that exists about a local business. Name, category, address, hours, phone, services, description, photos, and every review attached to it. It is the closest thing you have to an official record.

Your website. Mostly what you say you do and where you do it. Worth noting that a site which never names the services in plain words, or never names the town, is harder to place than owners assume.

Anywhere else your name appears in public. Directory listings, industry sites, local news, chamber pages, social profiles, forum threads where somebody asked for a recommendation and people answered.

Reviews, everywhere they live. Google especially, because there is more of it than anything else, but not only Google.

Notice what is not on that list. Your ads are not on it. Your mailers are not on it. The thing you tell people on the phone is not on it. Only what is written down in public counts.

Two businesses, same street

Here is the part worth sitting with. Take two shops in the same category a block apart. Same quality of work, same prices, same everything a customer would care about.

The first has a filled out profile, services named the way customers say them, photos from this year, and a steady flow of reviews where people describe what was actually done for them.

The second has a category from setup day, no description, four photos from when it opened, and nine reviews, the newest one from two years ago.

If you had to write one sentence recommending one of them, which one could you even do it for? You could describe the first one in detail. The second one you could barely confirm is still open.

That is the whole thing. There is no algorithm trick underneath it. One business is easy to describe and the other is not, so one gets named and the other does not get mentioned.

The word thing matters more than people think

There is a specific version of this that costs businesses recommendations quietly.

Customers ask in their own words. They ask for someone honest, someone who will not upsell them, someone open on a Sunday, someone good with anxious dogs, someone who does the small jobs.

Those are not industry words. They are customer words. And they live in two places: your own listed services and descriptions, and your reviews.

You control the first one directly. Go list your services in the language a customer would say out loud, not the language on your invoice software. The second one you influence by having enough recent reviews that your customers' own vocabulary is sitting there in volume.

Do this

  1. 1Ask an AI assistant for the best businesses in your category and city, then ask it what it knows about yours by name. Read both answers for what is missing or wrong.
  2. 2Open your Google Business Profile and read it like a stranger. If it does not say what you do, where, and when, that is the first fix and it is free.
  3. 3Rewrite your services list in customer words. "Brake repair" not "undercarriage service." "Same day" not "expedited scheduling."
  4. 4Put current photos up. A profile whose newest photo is years old reads as a business that might not be there anymore.
  5. 5Count your reviews and check the date on the newest one, then do the same for the two competitors that got named instead of you.
  6. 6Set up a way to ask every customer that takes two seconds, because the ones that take a conversation do not survive a busy week.

You cannot edit the answer, only the sources

There is no login where you correct what an assistant says about your business. That is worth accepting early because it saves you looking for one.

What you can do is change what is publicly true and publicly visible, and then it works itself out. Every one of these systems is downstream of the same material, which means fixing that material fixes all of them at once, including the plain old search results and the map pack you were already trying to win.

Reviews are the piece of it that never stops needing attention, and they are the piece almost everybody lets slide, because asking is awkward and remembering to ask is its own small job. Take the friction out and the problem mostly solves itself. A tap card at the counter turns the ask into a two second thing. Have a look at shop.optechsol.llc.

Nothing on that list is clever. It is just being the business that is easy to vouch for, and most of your competitors have not done it either.