Why ChatGPT Picks One Plumber and Never Mentions Yours

The short answer

When someone asks ChatGPT or Gemini "who's the best HVAC company in [city]," the AI doesn't rank ten options and let the customer pick. It picks one, maybe two, and moves on. That's the whole problem. You're not competing to be in a top ten list anymore. You're competing to be the one name that gets said out loud.

Why it only gives one answer

Think about how you'd answer if a neighbor asked you who to call for a cracked heat exchanger. You wouldn't read them a list. You'd say a name. AI models are built to answer the way a person would — direct, confident, one recommendation. That's the whole point of the product. Nobody wants to ask Gemini a question and get a spreadsheet back.

So the model has to choose. And it chooses based on whatever it can find that looks like agreement. If five different sources — your website, Google reviews, a local Reddit thread, a "best of" roundup on a news site — all point to the same business for the same kind of job, the model treats that as settled. It's not doing math. It's pattern-matching on which name keeps showing up attached to the question being asked.

This is the part most owners don't expect: the model isn't necessarily picking the best plumber. It's picking the plumber it can most confidently describe. Confidence comes from repetition and clarity, not from being the best crew in town.

Why your competitor keeps winning

We've looked at this across HVAC, plumbing, and electrical accounts, and the pattern is almost always the same. The business that gets named isn't spending more on ads. It's not necessarily bigger. It just has a cleaner footprint:

• Its name, services, and service area are described the same way everywhere — website, Google Business Profile, directories, review sites. No version says "HVAC repair," another says "heating and cooling," another says "climate control services."
• It has reviews that mention specific things — "fixed our furnace on a Sunday," "replaced the water heater same day" — not just star ratings. Specific language in reviews gives the model something concrete to quote or paraphrase.
• It shows up answering the actual question people ask, not just ranking for generic keywords. "Emergency plumber open now" and "best plumber in town" are different questions with different right answers, and the model treats them differently.
• Nothing contradicts it. No old directory listing with a disconnected number, no NAP mismatch, no review that says "used to be good, went downhill."

Meanwhile the business that doesn't get named usually has scattered information. Old address on Yelp. Different phone number on a directory from 2019. Services described vaguely. Reviews that are all just stars, no text. None of that is disqualifying on its own — but it means the model has less to go on, and less confidence, so it defaults to whoever it's more sure about.

What we don't know for certain

We'll be straight with you: nobody outside OpenAI, Google, and Anthropic knows the exact weighting of their models. We don't know if a review from three days ago counts more than one from three years ago, or by how much. We don't know if Reddit threads carry more weight than directory listings for local trade searches specifically. Anyone who tells you they've reverse-engineered the exact algorithm is selling you something.

What we do know, from watching real accounts across markets, is the pattern above holds up. Consistency and specificity beat volume. A business with fifty scattered, contradictory mentions loses to a business with fifteen consistent ones.

What actually changes the outcome

You can't buy your way into being the AI's answer the way you can buy a top ad slot. But you can clean up the things that make a model unsure about you:

• Fix every listing so your name, address, phone, and service area match exactly, everywhere.
• Write your service pages in the plain language customers actually search — "AC not cooling," "burst pipe," "breaker keeps tripping" — not just industry terms.
• Get reviews that describe the actual job, not just "great service." Ask customers to mention what you fixed.
• Check what's already being said about you on sites you don't control. A stale directory or a bad old review can be quietly working against you.

None of this is exotic. It's the same blocking-and-tackling that's always mattered for local search — it just matters in a new place now, and the stakes are higher because there's no page two. There's just the one name the AI says, and then silence.

Where to start

Most owners have no idea what ChatGPT or Gemini actually says when someone in their market asks "who's the best." They assume they show up because they show up on Google. Those are two different questions with two different answers, and the gap between them is usually where business is quietly leaking out.

Run a free scan at hollar.ai/scan and see exactly what the AI models say about your business right now — and who they're naming instead.