How to structure projects and monitors for a multi-location brand
Workspace design for tracking a multi-location brand in AI search: when to use one project or several, one monitor per location, tags per service line, and the response budget math that decides the plan.
A multi-location brand can set up AI visibility tracking in an afternoon and spend the next six months regretting the structure. Locations get mixed into one monitor, service lines can't be separated, and the response budget runs out in week three. The fix is to decide the structure before you add a single prompt.
This guide uses Promptwatch's building blocks (projects, monitors, prompts, tags and topics) because that's the tool we'd use for it, but the reasoning carries over to most platforms.
One project or several
Our rule is one brand, one project, treating the project as the home for that brand's prompts, monitors, competitors and website data. If all forty branches trade under the same name on the same website, they belong in a single project, because the crawler logs and AI-referred visits for that website are one data set. Splitting them would scatter the information you need when you trace a weak location back to its page.
Use separate projects when the things you're tracking are actually separate. A group that runs two brands on two websites needs two projects. So does a brand with a distinct site per country, where each site has its own pages and its own crawl data. An agency handling several clients gives each client a project.
One monitor per location
In Promptwatch, location is set on the monitor: a country, an optional state and an optional city, plus a language, an optional persona and the models to query. That makes the structure simple. Each location you want to measure gets its own monitor, and every monitor carries the same prompt panel.
Name monitors consistently, for example "UK / Manchester / EN", so that sorting and filtering work later. Keep the prompt panel identical across monitors. If one city needs a prompt the others don't, add it to all of them or to none. The moment wording differs between locations, the comparison stops meaning anything. How to write local prompts covers writing a panel that works everywhere.
For branches inside one big city, a city monitor may be too coarse. Coordinate targeting lets a monitor ask from an exact point with a radius from 20 m to 5 km. ChatGPT, Google AI Overviews and Google AI Mode answer as if asked from there; the other models use the country, state and city. It's available on Business, Agency Scale and Enterprise.
Tags and topics per service line
Most multi-location brands sell more than one thing. A dental group might do check-ups, implants and orthodontics, and visibility for one tells you little about the others. Group prompts by service line with topics or tags, so the report can show "implants in Leeds" separately from "check-ups in Leeds".
Promptwatch also lets you set a prompt type (organic, brand, competitor comparison) and an intent on each prompt. Use them. A branch that's named for branded questions but never for unbranded ones has a different problem from one that's missing everywhere.
Do the response budget math
Every check of one prompt on one model on one day is one response. The monthly total is:
prompts x models x days x locations
Take a hypothetical brand with 30 prompts, 4 models, daily runs (30 days) and 8 locations: 30 x 4 x 30 x 8 = 28,800 responses a month. Here's how that lands on Promptwatch plans:
| Plan | Price | Responses | City tracking | Fits? |
|---|---|---|---|---|
| Professional | $245/mo | 18,000 | No, country level | No city monitors |
| Business | $579/mo | 42,000 | Yes, plus coordinates | Yes |
| Agency Kick-off | $199/mo | 10,000 | Yes | No |
| Agency Growth | $399/mo | 25,000 | Yes | No |
| Agency Scale | $799/mo | 65,000 | Yes, plus coordinates | Yes |
Business also caps prompts at 350. If you count each prompt in each monitor separately, which is the cautious way to plan, this example uses 240 (30 x 8). Agency plans have unlimited prompts, so only the response total matters there.
The quickest way to fit a tighter budget is tiering. Give your top locations the full panel and the rest a short list of the prompts closest to a booking. Five locations at 30 prompts plus ten at 10 prompts, on 4 models daily, is (150 + 100) x 4 x 30 = 30,000 responses. Dropping a model you don't report on saves a quarter of that. How many cities to track goes deeper on tiers.
Plan choice
For a brand running its own tracking, Business is the starting point once you need city monitors. It's $579/mo with no free trial, and it comes with 5 projects and 5 seats, 42,000 responses, state and city tracking, coordinate targeting, Ads Radar, Shopping Insights, the automated Content Agent and an audit log. Professional ($245/mo, 2 projects) works only if country-level rows are enough, for example a brand with one site per country that doesn't need city detail.
Agencies running several multi-location clients should look at the agency plans. Every one has unlimited projects and prompts, 10 seats, city tracking, crawler logs, visitor analytics and the Looker Studio connector. Kick-off and Growth have a 7-day trial on monthly billing; Scale doesn't, but it's the agency plan with coordinate targeting. Agency plans don't include 360 Insights or the Content Agent, so a brand that wants Search Console data overlaid on AI citations, or scheduled publishing, is better served by Business. We compared agency options in best GEO tools for agencies with multi-location clients.
Reporting the structure back
A good structure pays off in the report. Show one row per location per engine, never a single averaged score. For each weak location, follow it through: who got named instead, which pages the answer cited (citation analytics, with Page Tracker on the branch page), whether AI crawlers fetched that page (Agent Analytics, from Professional and on all agency plans), and whether AI-referred visitors to it converted (visitor analytics). The Action Board turns those findings into a prioritized to-do list, and Agent Chat can build the location dashboard or a PDF report from the same data.
That chain, from a city's score to the page behind it, is why we recommend Promptwatch for multi-location brands. The prompt tracking is table stakes. What's harder to find elsewhere is the crawler and visitor data for each location page in the same workspace. Plans and limits are on promptwatch.com.