How to track "near me" prompts in AI search
"Near me" prompts carry no place name, so the answer depends entirely on where the AI engine thinks the user is. How to track them by country, city and exact coordinates, and why the city belongs in the monitor, not the prompt.
A prompt like "dentist near me" has no place in it. Everything that makes the answer local comes from outside the text: where the engine believes the person asking is. That makes "near me" the purest test of local AI visibility you can run. It's also the easiest one to measure wrong.
The usual mistake is checking on your own phone. Your ChatGPT has some idea where you are, so you see the answer for your office, today, once. Customers ask from home, from the train, from the other side of town. Tracking "near me" properly means controlling the location the engine is given, then asking the same question every day.
Why the wording changes the measurement
Compare three prompts: "best Thai restaurant", "best Thai restaurant in Leeds" and "best Thai restaurant near me". The first is ambiguous, and an engine may answer with general advice or ask where you are. The second puts the location in the text. The third asks the engine to supply the location from context.
These are different measurements. With the city typed in, a tracker mostly checks whether the engine knows the Thai restaurants in Leeds. With "near me", it checks whether the engine connects the user's location to you. A business can do well on one and be missing from the other, so track them as separate prompts and don't add them together.
Three ways to set the location
A "near me" prompt is only as good as the location attached to it. In Promptwatch the location sits on the monitor, at one of three grains.
Country-level tracking is on every paid brand plan from Essential ($95/mo) and on every agency plan. It's too coarse for most "near me" work, since an answer for a whole country says little about any one shop.
City-level tracking is on Business ($579/mo) and every agency plan, starting with Kick-off at $199/mo. The monitor tells the engine the user is in a given city, with a state field for US monitors. This grain applies to every engine Promptwatch tracks, including Perplexity, Gemini and Copilot.
Coordinates are the finest grain. Local GEO/SEO targeting, added in October 2026, pins the monitor to an exact point with a radius between 20 m and 5 km. ChatGPT, Google AI Overviews and Google AI Mode then answer as if asked from that spot, and Promptwatch describes it as the way to see "near me" visibility per location. It's on Business, Agency Scale ($799/mo) and Enterprise. Other models in the same monitor fall back to the monitor's country, state and city.
For a single-location business, the setup that works is usually a coordinate monitor for the three engines that honor it plus a city monitor for the rest. For a business that serves a whole city, a city monitor may be all you need.
Keep the city out of the prompt
When the monitor sets the location, write the prompt without one. "Emergency plumber near me" in a monitor pinned to your service area is a clean test. "Emergency plumber near me in Bristol" is muddier, because you've given the engine the place twice and can no longer tell which signal it used.
There's a practical reason too. A prompt without a place name can go into every location's monitor unchanged. Promptwatch lets you copy prompts into other monitors, so you write the set once and every location answers exactly the same question. Comparisons between locations only hold when the wording is identical.
If you want to know whether typing the city changes the answer, test it on purpose. Put the "in Bristol" versions in their own monitor, tag them and compare. Just don't mix both styles in one monitor and read the result as one number.
Build the prompt set
Start from what customers would type. For most local categories that's a mix of plain "near me" discovery prompts ("vet near me", "vet open now near me"), the same need phrased without "near me" ("where can I get my dog vaccinated this week") and a few branded checks ("is [your name] open on Sundays").
Label them. Promptwatch has prompt types (organic, brand, competitor comparison) and free-form tags, so a "near-me" tag lets you filter results down to just those prompts later. Fifteen to thirty prompts per location is a reasonable start, and it's close to what Promptwatch itself suggests per monitor.
Read the raw answers before the score
For "near me" prompts, open the Responses tab before you look at Visibility Score. You're checking whether the engine actually answered locally. If an answer lists businesses across the city, gives generic advice or asks the user where they are, your monitor may be set at a coarser grain than the question needs. Write that down. Those answers still count against you in the score, but the fix is in the setup, not in the business.
Then compare grains. If ChatGPT at your coordinates names you and Perplexity at city level doesn't, that isn't necessarily an engine preference. The two answers were given for different locations. Our guide to tracking Perplexity, Gemini and Copilot for local queries covers how to make that comparison fair.
Give it two weeks of daily runs before reacting. Local answers move between days, and one run is an anecdote.
When competitors get named and you don't
Look at what the answers cite. Citation analytics in Promptwatch lists the pages and domains behind each answer, including directories and Reddit threads, plus offsite mentions that name a business without linking to it. If the "near me" answers keep citing a directory you're missing from, that's a listings job. If they cite a competitor's location page, compare it with yours.
Then check whether AI crawlers can reach your location page. Agent Analytics logs IP-verified visits from bots such as OAI-SearchBot, ChatGPT-User and PerplexityBot, with errors, on Professional and up and on every agency plan. A page that returns errors to those bots is hard for the engine to cite.
The setup we'd use
For one location: Promptwatch Business, one coordinate monitor with ChatGPT, AI Overviews and AI Mode, one city monitor for the other engines, and the same place-free prompt set copied into both. Agencies on Kick-off or Growth get the city half of that and need Agency Scale for coordinates. For the city side in more detail, see how to track ChatGPT recommendations by city, and for exact points, our coordinate setup guide.