How to choose a radius for local AI tracking

When you pin a Promptwatch monitor to exact coordinates, you also pick an accuracy radius: 20 m, 50 m, 100 m, 500 m, 1 km or 5 km. ChatGPT, Google AI Overviews and Google AI Mode then answer as if the prompt was asked from that spot.

We haven't seen a published study of how each radius changes AI answers, and we haven't run one we'd stand behind. So this guide won't tell you "use 500 m". It gives the reasoning for each type of business and a way to test your choice in about two weeks.

What the radius is for

The pin says where the user is. The radius says how precisely. A 20 m radius describes someone standing outside a specific door. A 5 km radius describes someone somewhere in a district.

Neither is more correct in general. The right radius matches how uncertain you are about where your real customers are when they ask. Two questions get you most of the way: where do customers ask from, and how far will they travel?

The radius is fixed once the monitor exists, along with the point. Changing it means creating a new monitor, so the choice is worth a few minutes of thought.

Walk-in shop

A café, a bakery, a phone repair counter. Customers are often close by when they ask, and they won't walk far. A tight radius (20 m to 100 m) on the storefront is a fair model of "someone on this street looking for coffee".

The catch: a pin on your own door may flatter you, if engines lean heavily on distance. We don't know how much they do, which is exactly why it's worth checking. Add a second monitor at a point a few streets away where customers come from, such as a station exit or an office block, with the same radius. If you're named at your door and missing two streets away, you've learned how small your AI catchment is.

Service-area business

A plumber, a cleaning company, a mobile vet. The address matters less because you go to the customer, and there's no single "near me" point. People ask from their homes across the area you cover.

Here a wider radius (1 km or 5 km), centered on the neighborhoods you serve, is the more honest setting. One pin on the depot tells you what someone at the depot hears. Several pins spread across the service area tell you what customers hear.

Five kilometers is the largest option. A service area much bigger than that needs several coordinate monitors, or a city-level monitor for the overall picture plus coordinate monitors for the areas that matter most.

Dense downtown

Where competitors sit on every block, like restaurants in a city center or clinics along one busy street, a short distance may change who counts as "nearby". That argues for a smaller radius (50 m to 500 m) and for more than one point. A 5 km circle in a dense center takes in far more competitors than any customer would consider on foot.

Suburban destination business

A dentist, a gym, a garden center. Customers drive a few minutes and usually ask from home. A 500 m or 1 km radius around a residential area you draw from fits that better than 20 m on your car park.

Local Falcon's radius is a different setting

If you've used Local Falcon, its radius sounds similar but plays a different role. A Local Falcon AI scan takes a location, a keyword, a grid size and a radius, and every grid point is a separate AI sample. Its knowledge base suggests a radius of 2 miles in cities, 5 in suburbs and 10 in rural areas. Promptwatch's radius describes precision around one point, and 5 km is its maximum. Don't copy a Local Falcon setting across. Our Promptwatch vs Local Falcon comparison covers the two approaches in more depth.

How to test a radius

If you're torn between two radii, run both. Because the radius can't change after creation, a test means two monitors:

Same point, same prompts (copy them between monitors), same language, same persona, created on the same day.
Only ChatGPT, AI Overviews and AI Mode. The other engines ignore the pin and use the monitor's country, state and city, so they add cost without telling you anything about the radius.
Two radii you'd actually consider, such as 100 m and 1 km.
Two weeks of daily runs before comparing.

Then compare the named businesses prompt by prompt. Promptwatch's competitor visibility and share of voice show who appears in each monitor, and the Responses tab shows the full answers. If the two lists are mostly the same, the radius doesn't matter much for this location. Keep one and pause the other. A paused monitor keeps its history, drops out of your dashboards and stops counting against your prompt quota. If the lists differ, read both sets of answers and decide which better describes where your customers are when they ask.

Budget check before you start: 20 prompts on three models is 60 responses a day per monitor, roughly 1,800 a month. A two-radius test for one location is about 3,600. Business includes 42,000 responses a month and Agency Scale 65,000.

What to record during the test

Keep notes per prompt, not just the headline score:

whether you're named in each monitor, and your Position in Answer
which competitors appear in one monitor and not the other
whether the cited sources change, using the Citations view for each monitor

A change in sources can be more useful than a change in names. If the wider radius pulls in a citywide directory and the tighter one cites a neighborhood guide, that tells you where to work, whichever radius you keep.

Where to run it

Coordinate targeting is on Promptwatch Business ($579/mo), Agency Scale ($799/mo) and Enterprise. It isn't on Essential, Professional, Kick-off or Growth, and neither Business nor Scale has a free trial. Once the radius is settled, the same monitor keeps feeding citation analytics and Agent Analytics, so you can see which local sources the answers rely on at that spot and whether AI crawlers reach your location page. The rest of the setup is in our coordinate tracking guide.