We Ran the Same Query in Three Towns - AI Visibility Swung 7x

Same business, same service, same exact set of questions. The only thing that changed between runs was the city name in the query and that alone was enough to swing AI visibility from 42.5% down to 6.25%. Here's what that tells us about how AI search actually treats "local."

The experiment

Take a home-care and aging-life-care business — the kind of company where "local" isn't a marketing angle, it's the whole model. Someone looking for elder care in Boston needs a provider who can actually show up in Boston, not a well-optimized website three states away.

So instead of running one generic visibility experiment, we ran the same query set — short-tail terms like "best elder care services" and "best geriatric care manager," plus long-tail, natural-language questions like "how do I choose a home care agency for an elderly parent?" — three separate times, once for each of three towns: a large core market, a smaller but established one, and a secondary market the business hadn't built much of a presence in yet. Same intent, same phrasing pattern, same models. The only variable was which city name got swapped into the query.

What we found

The business's visibility — how often it got mentioned at all across the response set — wasn't close to flat across the three towns.

Overall visibility across all three towns combined was 27%. But that single number hides the real story: in its biggest, most established market, it showed up in 42.5% of relevant responses. In the smaller but still-established town, that dropped to 32.5%. In the market it hadn't really built a presence in yet, visibility fell to 6.25% — about a seventh of the top city's rate, on the exact same questions.

Why this happens

Two things are going on underneath this.

First, the way AI models actually search hasn't changed just because a query is hyper-local: they still break the question down into background fan-out sub-queries before answering. That means the exact phrasing someone types matters less than whether the underlying intent — and the location — are actually reflected in indexable content somewhere.

Second, and more specifically for location: when we looked at what was actually being cited across these responses, the top sources weren't generic authority sites. They were location-specific. The single most-cited domain was the town's own .gov site, followed by aging-care directory sites — several of which had built out individual pages per city (a competitor page structured like citydomain.com/newton-ma, for instance). The business's own site was the 5th most-cited source overall, cited on the strength of its existing homepage, about page, and locations page — but it didn't have dedicated, individually-optimized pages for each town it operates in yet. The towns where competitors had that structure and this business didn't are exactly the towns where visibility dropped hardest.

In other words: the AI isn't just recalling "this company does elder care." For local queries, it's assembling an answer out of whichever sources are most specifically matched to that town — government pages, local directories, competitor city-pages — and a business without its own location-specific footprint simply has less to be cited from.

Worth noting: the type of sources being cited stayed fairly consistent across all three towns — roughly the same split between government sites, directories, and earned media in each market. What changed wasn't the shape of where the model was pulling from, but which specific pages within that shape actually existed for each town.

Is this just a local-business problem?

Yes - and this time we can say so with confidence, not just a hunch. The pattern holds specifically because location is part of what the person is actually asking for.

Think about the difference between "best massage places" and "best recruiting tool." The first is inherently local — wherever you are, "best" implicitly means "best near me," so the answer should and does change city to city. The second isn't location-bound at all — a recruiting tool doesn't care what city you're in, so there's nothing for location to change. Location matters when it's actually part of the customer's intent, and it stops mattering the moment a product or service isn't tied to a place.

That also explains why the type of source citied stays stable even as the specific businesses mentioned change: the model isn't treating "local" and "national" categories differently in terms of which kinds of sources it trusts — it's just that for a local-intent query, the most relevant version of those sources is the one built around that specific town.

What this means for you

If your business serves multiple, distinct local markets:

• Don't treat "AI visibility" as a single company-wide metric. Measure it per market. A strong overall number can hide a market where you're nearly invisible.

• Build location-specific pages, not just a locations list. The sources winning citations in underperforming markets weren't generic — they were pages built around that specific town.

• Check what's actually being cited in your weakest markets before you write anything new. The gap here wasn't really about content quality — it was about not having a page that matched the query's location as specifically as a competitor's did.

• If your product isn't location-bound, don't chase this. This effect is specific to businesses where "near me" is implicitly part of what people are asking. A broader product or SaaS tool won't see the same swing, because location was never part of the question.
  
  

How XLR8 AI fits into this

This kind of market-by-market breakdown — not just "are we visible," but "are we visible in the specific places our customers actually are" — is exactly what XLR8 AI's platform is built to surface, the same way it's helped us track shifts in fan-out query behavior and other AI-search patterns. If you run a multi-location business and want to know how your visibility actually breaks down by market, grab time for a walkthrough.

FAQ

Does AI search really give different answers based on location?

Yes — for queries where location is part of the intent, the sources an AI model pulls from (and therefore which businesses get mentioned) can shift substantially by city, even when the underlying question is identical.

Why did visibility drop so much in the smaller market?

Mainly because the sources getting cited there were more specifically matched to that town — a competitor's dedicated city page, a local directory, a government resource — and this business didn't yet have an equivalent page built specifically for that market.

Does this apply to national or non-local brands too?

No — and that's the key distinction. Location matters for AI visibility specifically when location is part of what the person is actually asking for, the way it is for a local service like elder care or "best massage places nearby." For a broader, non-location-bound product — a SaaS tool, for instance — where you're searching from is largely irrelevant, because it was never part of the question in the first place.

Does the type of source cited change based on location too?

Not much. Across the three towns, the overall mix of source types — government pages, directories, earned media — stayed fairly consistent. What changed was which specific pages existed for each town, not the categories of sources the model trusted.

What should a multi-location business actually do about this?

Measure visibility per market rather than as one blended number, and build a genuinely location-specific page for each market you care about rather than relying on a single locations list or your homepage to carry all of them.