How to measure local AI visibility for a service-area business
A how-to for plumbers, cleaners, mobile mechanics and other businesses that work in towns where they have no storefront: choosing which towns to track, city vs coordinate monitors, and what to put in the monthly report.
A service-area business is measured in the wrong place by default. Most local tools anchor to an address, and your address is wherever the van parks at night. The jobs come from the towns around it. When someone in one of those towns asks ChatGPT or Google's AI Mode for "a carpet cleaner who can come this week", what matters is the answer they get there.
Google's own Business Profile help reflects this. Its local ranking page asks for your full address only if customers can visit. So a business without a shopfront is normal. Measuring it just takes a different setup. We ranked the tools that can do it in best GEO software for service-area businesses. This guide is the method.
Choose the towns from your job records
Don't start from a map with a radius drawn on it. Start from where the money comes from. Pull the last twelve months of jobs or invoices and group them by town. You'll usually find a few towns doing most of the work, a middle group you'd like more of, and a long tail of one-off jobs.
Track the first two groups. The top towns are where losing AI visibility costs you now. The middle group is where you want to grow, and a competitor who's always named there is useful to know about. Leave the long tail out, or rotate one or two towns in each quarter as a spot check. How many cities to track in AI search has a tiering method that fits this well.
Then check whether location changes the answer in your trade at all. Run the same 15 prompts in your two most different towns for two weeks. If the businesses named and the sources cited barely differ, one monitor per region may do. If they differ, give each town its own.
City monitors or coordinate monitors
For most service-area businesses the town is the right unit. Set the city on a monitor, attach your prompt panel, and repeat for each town. The prompts stay identical; only the location changes. That way a difference between two towns is a real difference and not a wording artifact.
Coordinates are worth it in two situations. The first is a big city where you only serve some districts: a city-level monitor asks as someone in the city as a whole, while a coordinate monitor asks from the district you care about. The second is the edge of your service area, where you want to know whether you're still named or whether a business closer to that spot takes over. For Maps, Google lists distance from the searcher as one of the main local factors. It doesn't document how its AI features treat distance, which is why measuring from the exact spot beats assuming.
Write prompts the way customers ask
Service-area searches lean on urgency and fit: "emergency electrician near me", "end of tenancy cleaning", "mobile tyre fitting that comes to the house". Most won't name a town, so the monitor has to supply it. Add a few that do name the town, using one template and swapping only the name. Keep branded prompts in their own group. How to write local prompts has the full method.
Check the town pages
Most service-area businesses have a page per town, and those pages are often thin. Before rewriting them, find out which ones AI engines actually read. Crawler logs show whether bots such as OAI-SearchBot, ChatGPT-User, PerplexityBot and Claude-SearchBot request each town page and whether those requests fail. A town page that's never fetched can't be cited, however good the copy. For Google's AI Overviews and AI Mode, Google's AI features documentation says the page has to be indexed and eligible to show with a snippet, so check indexing in Search Console as well.
What to put in the monthly report
Keep it to what the owner can act on. Per town:
- Is the business named, per engine, and where in the answer?
- Who is named instead, and how often?
- Which sources did the answers cite: your town page, a directory, a forum, a competitor?
- Did AI crawlers fetch the town page this month, and did any requests error?
- How many AI-referred visits landed on that town page, and how many became enquiries?
Don't average the towns into one score. "Named in four of six towns" with the two missing towns listed is more useful than any single percentage. Each missing town turns into a short to-do: get onto the directory that's cited there, fix the page the bots can't reach, or add the details the cited competitor page has and yours doesn't.
Measuring it in Promptwatch
Promptwatch handles this without an address. Each monitor carries a country, an optional state and city, a language and the models to query, and runs daily. A business working twelve towns builds twelve monitors with one prompt panel.
City and state tracking start on Business ($579/mo, no trial), which also includes local GEO/SEO coordinate targeting with a radius from 20 m to 5 km. In a coordinate monitor, ChatGPT, Google AI Overviews and Google AI Mode answer as if asked from that point, and other models fall back to the city. Agencies get city tracking on every agency plan from Kick-off ($199/mo); coordinates are on Agency Scale ($799/mo). Essential and Professional are country level only.
The report items above map onto features. Citation analytics gives page- and domain-level citations per answer, and Page Tracker follows individual town pages. Agent Analytics logs IP-verified AI crawler visits, with stackable path filters so you can isolate a /areas/ folder, plus error explanations and a citation rate per page. Visitor analytics attributes AI-referred visits and conversions through a script, the GTM template or the WordPress plugin. The Action Board collects the fixes into a prioritized list.
For a sole trader in one town, that's more than you need. For a business whose revenue is spread across a dozen towns with no shopfront in any of them, it's the setup we'd use, because it measures each town and the page behind it in one place. Plans are on promptwatch.com.