Why ChatGPT recommends some plumbing companies and not others

Ask ChatGPT to recommend a plumber in your city and it will name two or three companies, usually with a sentence about each: who they are, what they handle, how their reviews look. Ask why those companies and not yours, and it won't really tell you. The selection looks like a black box.

It mostly isn't. The way these systems pick local businesses leaves a paper trail, and in 2026 enough research has piled up to describe the pattern plainly. Here is what it shows.

The answer set is tiny

Start with the scale of the thing. SOCi's 2026 local visibility index found that ChatGPT recommends only about 1.2% of local business locations. That is not a typo. The overwhelming majority of plumbing companies never appear in any AI answer, and the few that do appear collect recommendation after recommendation.

That sounds discouraging, but it cuts the other way too: the bar for joining a small club is specific, and most of your competitors have not read the research either.

Where the names come from

An AI assistant doesn't know your company the way it knows a fact from an encyclopedia. When someone asks for a plumber in a specific city, it goes looking, and what it finds determines what it says.

A 2026 research study that ran plumbing queries across all 50 states found that seven of the twelve most-cited domains in the answers were curated directories, not plumbing companies' own websites. The assistants lean on places that have already done the sorting: directory profiles, review platforms, and licensing records. Your website matters, but mostly as one corroborating source among several.

The practical translation: the AI is reading about you more than it is reading you. If your directory profiles are unclaimed, thin, or disagree with each other about your name, phone number, or hours, the systems doing the reading conclude you're not a safe recommendation.

The signals that show up over and over

The same 50-state study counted which trust signals appeared in AI answers about plumbers. A few patterns stood out.

Licensing led everything. License mentions appeared 411 times across the dataset, the single most frequent signal, and one major engine regularly cites the specific license class with a verification link. A printed license number is checkable against a state board, which makes it exactly the kind of fact a cautious system likes to repeat.

Reviews came next, and not the way most owners think. Star averages matter less than what the review text says. A review that reads "they replaced our water heater the same day and walked us through the permit" hands the machine a service, a speed, and a location in one sentence. Volume and recency matter as well; a profile that hasn't earned a review in eight months reads as dormant.

Longevity and service breadth rounded out the list. Years in business, clear service descriptions, emergency availability that's stated consistently everywhere. Mundane facts, machine-readable.

Notice what's missing: nothing in that list is exotic. There is no secret schema trick at the top. The systems reward the same boring trust infrastructure that a careful homeowner would check by hand, applied consistently across every place your business appears.

Why a competitor with worse reviews can outrank you

This is the question that actually stings. The answer is that the assistants weigh corroboration, not just quality. A shop with a 4.4 average that appears identically on six directories, prints its license number, and has review text full of specific jobs will beat a 4.9-star shop whose online presence is one Google profile and a Facebook page. The first company is easier to verify. To a machine deciding what to say out loud to a stranger, verifiable beats impressive.

What a small shop can do about it, in order

First, claim and reconcile your directory profiles. Same business name, same phone, same hours, everywhere. This is tedious and it is also the fix that pays back hardest on the list.

Second, put your license number on your website with the issuing state, not just the words "licensed and insured." Add it to your Google Business Profile and your main directory profiles too.

Third, change how you ask for reviews. Ask happy customers to mention the job: what you fixed, how fast you came. Twenty specific reviews teach the machine more than two hundred "great service" ones.

Fourth, make sure your website states plainly what you do and where. Service pages that name the work (drain cleaning, water heater replacement, sewer repair) and the cities you cover give the assistants text worth quoting.

What doesn't work: stuffing keywords into your business name (a fast way to lose your Google profile), paying for thin directory listings the engines don't read, or buying an "AI optimization" add-on before the basics above are done.