Does ChatGPT recommend your business?

Homeowners, renters, and businesses increasingly ask assistants who to hire the way they used to ask a neighbor. The assistant names three to six providers. If you're one of them, you get the call. If not, the customer never knew you existed — and nothing in your analytics will tell you it happened.

#1 in town. Invisible one town over.

In a real scan for a Glens Falls, NY electrician, the business ranked #1 for its home-city query and for EV-charger installation — its specialty. Five miles away in Queensbury, it appeared in 0% of answers; a competitor owned that shelf entirely.

That's the defining feature of local AI visibility: it is not one score, it's a map. Assistants form a separate shortlist for every town and every job type, and the boundaries don't follow your service area — they follow the model's associations. You can serve six towns and exist, as far as AI is concerned, in two of them.

This is not the Google local pack

Local SEO gives you a mental model — proximity, prominence, relevance — that only half-applies here. The differences that matter:

• No pin, no radius. The assistant isn't ranking by distance from a searcher. It's answering from what it associates with the town's name — which is why a service-area page listing twenty towns doesn't create presence in twenty answers.
• Ratings get quoted — and rounded, blended, or mis-attributed. Answers confidently cite review counts and stars. Sometimes they're stale or wrong. You want to know what's being said either way.
• Specialties are their own shortlists. “Electrician in Glens Falls” and “EV charger installation near Glens Falls” produce different shelves. A specialty you're known for can be your strongest AI asset — or a shelf you didn't know you'd already won.

What shapes a local AI shortlist

• Reviews, everywhere they live — Google, Yelp, Facebook, trade platforms. Volume, recency, and the words reviewers use (“panel upgrade”, “same-day”) all feed the association.
• Consistent name–address–phone across your site and directories. Conflicting listings blur the entity the model is trying to recommend.
• Your site saying the towns and the jobs in plain language. A page that names the service and the place ("EV charger installation in Queensbury") gives the model the association directly.
• Community mentions — local Reddit and Facebook threads, "who do you recommend" posts, local news. Assistants weight the same conversations neighbors do.

Works with the way you already run jobs

Run your schedule in Jobber, Housecall Pro or ServiceTitan? The towns in your job records are the towns we should be scanning — a service-area page lists twenty, your invoices say which six actually pay. Tell us which tool runs your business on the signup form; that answer decides what we connect next.

Measure the map before you work it

The improvement work above is standard local marketing — the missing piece is feedback. A credible read means asking the real questions (“best plumber in each town you serve”, plus your specialty jobs), repeatedly, from a logged-out baseline, and recording who the assistant names and in what order — the same method behind AI visibility measurement generally. That's what a Dizem snapshot does for a service business: give us your website and service area, and we show you which shortlists you make, town by town and job by job — and who owns the ones you don't.