More of your customers are asking AI before they ask anyone else. Not "where's the nearest plumber" on a map — actual questions, typed into ChatGPT or Gemini or Perplexity: who should I hire to redo my kitchen, is this company any good, who's the best HVAC outfit in town. The answer they get back shapes who they call. And most business owners have no idea what that answer says about them.
The good news is you can check it yourself. The catch is that most people check it wrong, ask the question that flatters them, see a nice answer, and walk away thinking they're fine. So before you go typing your company name into ChatGPT, here's how to test this in a way that tells you something true.
The question that feels good and teaches you nothing
The first thing almost everyone does is type in their own business name. "Tell me about Donnelly Heating & Cooling." And the AI says something pleasant — pulls from your website, your reviews, maybe your years in business — and you feel good and close the tab.
That tells you nothing. Of course it can describe your business when you hand it the name. The real question isn't whether AI knows you exist when prompted. It's whether AI brings you up on its own, when someone describes their problem and never mentions you at all. That's the moment that matters, and it's the one you have to test deliberately, because it won't happen by accident while you're poking around.
Write the prompt your customer would actually type
The skill here is writing prompts the way a real person with a real problem writes them. Nobody types your company name. They type their need.
Think about what someone in your area actually asks before they hire. For a heating and cooling company, it's things like:
- "Who's the best HVAC company in Westbrook?"
- "I need someone to replace my furnace in Westbrook — who should I call?"
- "Best heating and air conditioning repair in the Westbrook area"
- "Is Donnelly Heating any good?" (the direct one — worth testing, but it's the easy mode)
Notice the shape of those. They name a place and a need, not a business. That's buyer-intent — the language of someone about to spend money who hasn't decided on you yet. (And notice they say the town by name. More on why that matters than "near me" below.) Those are the prompts where you either show up or you don't, and where showing up actually means something.
Write five or six of these for your own business. Use the words your customers use, not your industry's jargon. Homeowners say "fix my AC," not "HVAC remediation." Get the phrasing right and the test gets honest.
Vary the inputs, because the answer moves
One prompt on one platform is an anecdote, not a result. A few things change the answer, and you want to see across all of them:
Name the place — don't lean on "near me." This one trips people up. When you type "best HVAC company near me," the AI only knows where "me" is if it has your location, and that depends on settings most people never turned on. On its own it's guessing from your IP address — so it might quietly turn "near me" into "near Boston" because that's where your internet provider looks like it is, and you'd never know it swapped your town for a city an hour away. The fix is simple: say the town. "Best HVAC company in Westbrook" is a controlled test you can trust and repeat. "Near me" is testing your own phone's settings as much as your business. Run it with your town named, then try the county and the nearest city too — the AI treats "Westbrook" and "the Worcester area" as different questions and may name a different set of businesses for each.
The platform. ChatGPT, Claude, Gemini, Perplexity, and Google's AI answers don't share a brain. They pull from different places and weight them differently. You can be the first name on one and invisible on another. Check all five — the one you're strong on isn't necessarily the one your customer is using.
Repetition. Ask the same prompt two or three times. These systems are a little different each time, and a business that shows up once out of three is in a weaker spot than one that shows up every time. You're looking for consistency, not a lucky single hit.
Read the answer honestly
When the response comes back, don't just scan for your name and call it a day. Look at:
- Are you named at all? Yes or no, before anything else.
- Where in the list? First is very different from sixth. People act on the top of the list.
- Who's ahead of you? Those are your real competitors in the channel that's quietly steering your customers — and some of them may surprise you.
- What does it say about you? Being mentioned dismissively is not the same as being recommended.
Here's the uncomfortable part, and the reason this is worth doing. Donnelly Heating & Cooling — to keep using the example — might have four hundred five-star reviews, twenty years in town, and a Google Ads budget. And you can still ask "best HVAC company in Westbrook," watch the AI name three competitors, and find Donnelly nowhere in the answer. Being good and being visible to AI are two different things. The first doesn't guarantee the second, and a lot of genuinely good local businesses are finding that out the hard way right now.
(To be clear, Donnelly's made up — Westbrook too. But the gap is real, and I've seen it on businesses that would recognize themselves in that description.)
Do this once. Then notice what it would take to keep doing it.
Run the test. Write your buyer-intent prompts, run them across the five platforms, vary the location, repeat each a few times, and write down where you land. It's genuinely worth an afternoon, and you'll learn something most of your competitors haven't bothered to find out about themselves.
Then notice what you just did. Six prompts, five platforms, a few repetitions, a few location variants — that's well over a hundred individual checks, and it's a snapshot of one day. AI answers shift week to week. To actually manage your visibility instead of spot-checking it, you'd be doing that again and again, tracking what changed and who moved ahead of you.
That part — running it systematically, every week, across every platform, and watching the trend — is the part that doesn't scale by hand. It's exactly what we built LLMClarity to do, and if you'd rather see your results than spend the afternoon gathering them, that's the reason it exists. But do the manual version first. Even one honest pass will tell you whether this is something you need to pay attention to. For a lot of businesses, it is.