Create a brand and branch register
Record the brand name, aliases and domain, plus separate location identifiers: city, name, address, phone, URL and genuinely available services. Distinguish a physical branch from a service area without premises. The register is a reference for judging answers, not a task to create new city pages. If AI names the chain without identifying a branch, count a brand mention but do not automatically assign a correct branch recommendation.
Build comparable city prompts
Keep the service and intent consistent while changing the city, for example asking for an installation provider with a defined service scope. Separate business discovery, comparisons and questions about a named branch. Match requirements to each location's real offering; identical questions about a service unavailable in one city are not a fair comparison. Select local competitors for each market instead of copying one national list.
AI monitoring budget: choosing questions, systems and review frequency
A city in the question is not geolocation
An explicit city name and a 'near me' question create different conditions. For the latter, record the actual location context if the environment provides it. Do not label a result geolocated simply because the prompt contains a city. Document the AI system or mode, language, country, date and controllable settings. If the tool does not support device location, state that limitation in the report.
Check if AI sees your business
Run a free audit
Score the brand, branch and facts separately
For every answer, flag a brand mention, recommendation of the correct branch, correct address or URL and service accuracy. Record cited sources and open them to confirm the location association. Recommending a branch in another city or a nonexistent location is an error even if the chain name is correct. Separately mark no answer, measurement failure and the absence of an AI Overview. These are different events.
AI citation audit: do sources really support information about your company?
Does the assistant send each customer to the right branch?
Imagine a key-cutting business with two branches. North Street cuts standard keys on Saturdays. Market Street is the only branch that programs car keys. Keep a reference sheet for each location with its address, services, hours and the public page that confirms them.
Test the same car-key request three ways: name the North Street area, name the Market Street area, then leave out the location. Record the branch recommended and the link a customer would follow. If the answer sends a driver to North Street for programming, count the brand mention and flag the service mismatch separately.
Treat a request for the customer's location as a distinct outcome on the location-free question. It may be the most useful response. Keep these general questions in their own group so they do not inflate or dilute either branch's result.
Compare without mixing samples
Hypothetical example: in 20 valid answers for city A, the correct branch is recommended four times. The result is 4/20, or 20%, only within that sample and environment. Report city B's own numerator and denominator. Do not combine different services or branded questions with category questions. For a network result, explain city weights: an equal-weight branch average answers a different question from a result weighted by test counts.
Correct branch recommendation: City A 4/20 = 20%; City B 6/60 = 10%. (20% + 10%) / 2 = 15%; (4 + 6) / (20 + 60) = 12.5%. Hypothetical data. Results describe this sample, not the entire market.
Check if AI sees your business
Run a free audit
Turn a local error into a specific task
If a source contains an old address or outdated opening hours, assign the correction to the data owner and check the branch page and relevant external profiles. If AI attributes a service to the wrong branch, examine how clearly its scope is described on the existing page. Gather missing facts from the client before editing. Do not mass-create city pages based solely on omissions. Record the changed URL, date and correction, then repeat a comparable series.
Did updating your article improve AI visibility? How to plan a before-and-after test
Organising Semly measurements and FAQ
In Semly, prompts, answers, competitors and sources can support analysis of local questions; options depend on configuration and plan. Name groups by city and service and retain these labels in exports. Does one brand appearance imply visibility for every branch? No. Does an improved score guarantee branch visits? No; contacts and sales require separate measurement. Can cities with different offerings be compared? Compare the shared scope and report additional services separately.



