Multi-location AI visibility: why AI recommends a different business in every city

Run “best HVAC repair near me” through ChatGPT from your office and it names your company. Run the same query from a suburb 20 miles away where you opened a second crew last year, and a competitor you have never heard of takes the top spot. Same brand, same query, two different answers. That gap is local AI visibility, and most multi-location businesses never see it because they only ever check one city: their own.

This is not a quirk. It is how the models work for local queries. When someone asks an AI assistant for the best plumber, dentist, or law firm in a place, the model leans on that place’s local signals: the Google Business Profiles, reviews, directories, and roundups that exist for that specific market. Those signals are different in every city, so your visibility is different in every city too.

The shift is already large. In a study of 540 local queries across three US cities and six industries, Whitespark found that Google AI Overviews appeared for 68% of local searches on average, against 39% for the traditional local pack. For a multi-location brand, the AI answer is now the front door in most of your markets, and you are graded separately at each one. This post is for the two people who feel that most: the agency running a franchise or multi-location client, and the owner who serves a handful of towns and assumes the one they live in speaks for all of them.

Why AI recommends a different business in every city

Traditional SEO taught everyone to think about ranking as a property of a page. Your page ranks or it does not. Local AI works differently, because the query is about a place, and the model has to decide who belongs in that place before it answers.

The mechanics are not a black box. AI assistants do not hold a private opinion of your company. They build an answer from what the web says about your business in a given market. As the team at Birdeye puts it, “LLMs don’t check a directory. They read everything that has been written about your business and synthesize a picture of each location from that.” The words that matter are “each location.” The picture is built per city, from per-city material.

That material is uneven by nature. Your headquarters has been open the longest, so it has the most reviews, the fullest Google Business Profile, and the most mentions in local press and “best of” roundups. The town you expanded into 14 months ago has a fraction of all of it. When the model builds its picture of the new market, there is simply less of you to find and more of whoever has been there longer.

The signal mix behind those local results is well documented. In Whitespark’s 2026 Local Search Ranking Factors study, Google Business Profile signals account for 32% of local pack ranking factors, with reviews and local links making up much of the rest. BrightLocal’s read of the same survey puts reviews at 20% of the local pack and links at 24% of local organic. Every one of those inputs is location-specific. A brand strong on them in its home city can be close to absent on them three towns over, and the AI answer reflects the gap.

AI is now the local-search layer, not a side channel

It would be easier to ignore all this if AI answers were a niche. They are not. Uberall reported in August 2025 that almost one in five consumers now use AI to find local businesses. That is a fifth of your prospects, today, getting an answer you may never see.

The surface is also everywhere for the queries buyers actually ask. The Whitespark data splits by intent: AI Overviews appeared for only 15% of bare local-intent queries like “tacos san francisco,” but for 92% of informational queries like “how long does an eye exam take near me?” and 97% of hybrid queries. The higher-consideration queries, where a customer is comparing options and deciding, are exactly where AI dominates. Those are the ones worth real money.

A second large study points the same way and adds the geographic angle directly. Local Falcon ran 60,000 real-world simulations across 4,423 businesses in 20 countries and found that service queries like “fitness classes near me” are more geographically sensitive than broad “best of” queries in dense metros. Businesses closer to the searcher appeared in AI Overview results 72.0% of the time versus 68.5% for those slightly farther away, even inside a four-mile radius. Where the customer is standing changes who shows up. Spread that across the towns you serve and the gap gets wide.

One honest caveat: Local Falcon also found that AI Overviews do not use the local pack’s distance-ranking algorithm, so proximity affects whether you appear, not the order you appear in. For a multi-location brand, the point is not “get closer.” It is that the answer is computed fresh for each location, so it has to be measured fresh for each location.

The headquarters check is a false positive

I see the same failure repeatedly when I talk to local businesses. The owner hears AI search matters, opens ChatGPT, types their main buying query, and reads their own company name in the answer. Relief. “We’re fine on AI.” They close the tab.

The check was real. The conclusion was wrong. They tested their strongest market and read it as a brand-wide grade. HQ is where you have the deepest footprint, so it is the one city most likely to return your name. It tells you almost nothing about the markets where you are newer and thinner, which are usually the exact markets you are trying to grow.

The cost is quiet, which is what makes it dangerous. There is no error message when AI recommends a competitor in your satellite city. The lead just never contacts you. You do not see a lost form fill or a missed call, because from your side nothing happened. The customer asked, got a name that was not yours, and called someone else. A single HQ check will never catch it, and a brand-level dashboard that averages all your cities into one healthy-looking score can bury it just as well.

This is the same invisibility problem we cover in why your website is invisible to AI, with a local twist: you are not invisible everywhere. You are invisible in specific places, and the one place you keep checking is the place you are visible.

Five towns means five visibility positions

The reframe is simple to say and it changes how you work. Serve five towns and you do not have a visibility score. You have five of them. The table below shows what that can look like for an illustrative multi-location service business. The cities and numbers are made up to show the shape of the problem, not real results.

Market	Role	AI visibility (illustrative)	What AI tends to do for “best [service] near me”
Riverton	Headquarters	78%	Names you first or second, cites your reviews
Oakdale	Established (4 yrs)	54%	Names you mid-list, rarely leads with you
Castleton	Newer (18 mo)	22%	Names a competitor, you appear occasionally
Pinehurst	Newest (8 mo)	9%	Hands the answer to two competitors, you are absent
Brookfield	Expanding into	4%	Does not know you operate there at all

Read the HQ row alone and the business looks healthy. Read the whole column and the real story shows up: strong where it started, nearly invisible where it is trying to grow. The 78% in Riverton was never the number that mattered. The 4% in Brookfield is, because that is where the marketing budget and the new crew are pointed.

For an agency, this is the more useful way to report. A single “your AI visibility is 41%” line to a franchise client is an average that satisfies no one and points to no action. A per-city breakdown turns the same audit into a map of where to spend: which locations need reviews, which need local citations, which need to get onto the roundups AI keeps quoting. That is the difference between a number and a plan, and it is part of how we frame pitching GEO services to clients.

How to test local AI visibility per city

The fix for a blind spot is to look where you have not been looking. Testing per-city local AI visibility is not complicated, but it takes more than one query from your desk. Here is the approach.

• List the cities you actually serve. Not your legal address, your service area. Every town, suburb, or metro where you take customers is its own market with its own answer.
• Write the buying queries a real customer asks. “Best [service] in [city],” “[service] near me,” “who does emergency [service] in [area],” the comparison and decision queries, not just your brand name. These are what send a ready buyer to a competitor.
• Run each query for each city, several times. AI answers are non-deterministic, so the same prompt run twice can return different businesses. One run is an anecdote. Several give you a rate: “named in 6 of 10 runs in Castleton” is real signal.
• Weight toward where the customers are. Your primary market deserves the most runs. Adjacent markets get tested in rotation. Put the most data where the most revenue is, and a baseline everywhere you operate.
• Test the engines that matter, not just one. ChatGPT, Perplexity, Gemini, and Copilot, plus Google AI Overviews as a surface. The same query often returns a different shortlist on each, so a single-engine check is its own kind of HQ check.

This is the methodology BlueJar is built around for local businesses. When the brand is local, the audit auto-generates the catchment cities, weights the primary city heavily, and round-robins the adjacent ones, running 120 prompts per engine so the local-intent queries get real coverage in each market instead of one spot check in the main one. The output is a per-city view, which is the only view that tells a multi-location brand the truth. If you want the scoring model behind it, we explain it in what a GEO score measures.

What to fix once you can see it per market

Seeing the gap is most of the battle, because it tells you the fixes are local, not global. A new homepage will not make AI recommend you in Brookfield. The work happens market by market, on the signals that build the model’s picture of each place.

• Fill in the per-location Google Business Profile. A complete, correctly categorized profile for the weak market, not just the flagship. This is the single highest-weight local signal in the ranking-factor data.
• Earn reviews in the thin cities. The new location has fewer reviews because it is new. That is the most fixable gap, and reviews carry real weight in what feeds local results.
• Get into the local roundups AI quotes. The “best [service] in [city]” listicles, the directories, the local-press mentions. The model reads these per city, so being on the HQ-city roundup does nothing for the satellite city.
• Keep your NAP consistent everywhere. BrightLocal found businesses with consistent name, address, and phone data across citation sources are 40% more likely to appear in the local pack. Inconsistent listings in a new market actively hold it back.
• Build a real location page per city. One thin page that swaps the city name is not enough. Each market needs its own page with local detail, the way every home-services GEO playbook for multi-area businesses recommends.

For agencies, the same list is a sequenced service. Audit per city, find the two or three markets where the client is bleeding, and work those first. It is a cleaner story than “let’s improve your AI visibility,” and it maps to real locations the client cares about. That structure fits the work we describe in GEO for digital marketing agencies and how local answers get built on Google’s AI Overviews.

Want your real per-city numbers instead of an HQ guess? Run a free audit at bluejar.ai. For a local business it auto-generates your catchment cities and tests your buying queries across ChatGPT, Perplexity, Gemini, Copilot, and Google AI Overviews, so you see where you win and where AI hands the customer to someone else.