Local AI search playbook for home services companies
How we run local AI visibility for plumbers, HVAC firms, electricians and roofers: tracking each service-area town, emergency and near-me prompts, Ads Radar for sponsored answers, and town pages built only from real facts.
Home services companies have an awkward shape for local search. The business is registered at one address, maybe a yard or a small office, but the money comes from homes in a dozen towns around it. Customers rarely visit. They call, often in a hurry, and more of them now ask an AI engine who to call before they search anywhere else.
This is the playbook we run for plumbers, HVAC companies, electricians and roofers. Most of it applies to any trade that goes to the customer.
Track the towns you work in, not the address you're registered at
We start by asking the owner or the dispatcher where the jobs actually are. The answer is almost never "around the office." It's a list of towns, often with one or two the company wants more of.
On Promptwatch, the platform we run client programs on, each of those towns gets its own city monitor with the same prompt set. A city monitor sets the location the engines answer for, so we're measuring what a homeowner in that town sees, not what someone near the yard sees. City targeting is on every agency plan and on the Business brand plan; Essential and Professional track by country only.
Coordinate monitors have a place too, just a smaller one. They pin the location to an exact point with a radius between 20 m and 5 km, so ChatGPT, Google AI Overviews and Google AI Mode answer as if asked from there. For a home services firm we'd use one for a specific neighbourhood the company is targeting, not for the yard. Coordinate targeting needs Business, Agency Scale or Enterprise.
Our general onboarding sequence is in how we set up local GEO tracking for a new client.
Prompts that match how people ask in a hurry
Home services prompts fall into a few groups, and we cover each one in the set.
Emergency prompts are the ones that matter most for most trades. A burst pipe, no heat in winter, a power outage in part of the house, a leak after a storm. People phrase these as problems, not as service names: "water coming through my ceiling, who can come today" rather than "emergency plumber."
"Near me" prompts. We write one for each main service, because this wording depends entirely on where the engine thinks the user is. That's exactly what the city monitors are for. We leave the town name out of these, since the monitor already sets it.
Planned-work prompts: replacing a boiler or furnace, rewiring an older house, a roof replacement. These are longer decisions, and people ask about cost, how to choose a contractor, and what to check before hiring one.
Brand and comparison prompts: does this company cover a given town, is it licensed, how does it compare with the competitor the dispatcher hears about most.
We check prompt volumes and difficulty in Promptwatch before freezing the set and drop prompts nobody asks. For seasonal trades, prompt trends over time show when heating or cooling questions pick up, which helps schedule content before the season rather than during it.
Ads Radar: who is paying to appear in the answer
Organic visibility is only part of the picture once sponsored placements appear inside AI answers. Promptwatch's Ads Radar captures sponsored ads that appear in ChatGPT Search, Google AI Overviews and Google AI Mode answers to the prompts we track. For each ad it records the title, description, image or price, the advertiser's name and domain, the target URL and the prompt that triggered it.
For a home services client, the By Prompt view is the one we use most. It shows how many ads each tracked prompt triggered, next to the client's average visibility on that prompt. An emergency prompt where the client is barely named and three competitors are advertising is a different problem from one where the client is the first recommendation and nobody is advertising. The Top advertisers view tells us which competitors are spending, and the ad copy and landing pages show what they're promising.
Two limits worth knowing. Ads Radar doesn't cover Perplexity. And it doesn't report on a client's own OpenAI ad campaigns; that's what OpenAI's Ads Manager is for. Ads Radar is on Business and every agency plan. Our wider notes are in ChatGPT Shopping insights and Ads Radar.
Town pages with real facts only
The usual local SEO move for a service-area business is a page per town. We do that too, but only for towns where the company can say something true and specific. A page that swaps the town name into a template doesn't help anyone, and we won't publish one.
What goes on a real town page, if the client can supply it:
- which services the company actually offers in that town, and from which base
- the licences or certifications that apply there
- photos of jobs done in that town, with the customer's permission
- the response time the company can actually commit to, if it commits to one
- reviews from customers in that town, quoted accurately and with permission
If a town has none of that yet, it doesn't get a page yet. It stays on the monitor list so we can see what answers say about it, and the page waits until there's something real to write.
Each page gets LocalBusiness markup that matches the visible text. For a service-area business that means the real business address, and the markup only lists services the page lists. Our approach is in local business schema for AI search.
Check that the pages are read, then whether they bring calls
Once the town pages are live, we check Promptwatch's crawler logs for them. Which AI bots fetched each page (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and others), and did any of those visits hit errors? A new page that crawlers can't load won't be cited, and that's easy to miss without logs.
Then the citation data per town monitor: are the new pages being cited, and which local sites keep showing up behind the answers instead? That's where listing work comes from, and we cover it in how we find which local directories AI cites.
Last, visitor analytics. Promptwatch tracks AI-referred visits and conversions through a script, the Google Tag Manager template or the WordPress plugin. For a home services client, we set the conversion to whatever the business actually counts as a lead, usually a quote request or a booking form.
What we report
One row per town per engine, the emergency prompts broken out on their own, the ads appearing on tracked prompts, crawler errors on town pages, and AI-referred leads. No blended score across towns. The structure is in how we report local AI visibility to clients.
If you run a trade business and want to know what AI engines tell homeowners in your towns, send the town list to hello@1001seomedia.com.