Local AI Visibility Study: How Often AI Recommends Your Business

A local business that ranks well on Google Maps still faces a coin flip, one in two, when an AI assistant is asked for a recommendation in the same town. A study of 200,085 localized prompts run across ChatGPT, Google AI Mode, and Google AI Overviews found that the same prompt returns a different set of businesses about 70 to 80% of the time, even though individual businesses reappear in roughly half of follow-up runs. Google Maps, by contrast, mentioned the same tracked business 66% of the time and showed 99% agreement across repeated attempts, far steadier than any of the three AI surfaces tested.

What the study measured

Researchers analyzed 200,085 localized searches on three AI platforms: ChatGPT, Google AI Mode, and Google AI Overviews. Each search used a non-branded prompt a local business could plausibly be recommended for, drawn from a pool of 1,300 business locations. The team then ran the same prompt again and again over a 60-day window, and across multiple points on a map inside the same city, to see how much the returned business list moved around.

Two metrics did most of the work:

• Variance: how much the set of recommended businesses changed between runs.
• Persistence: how often a business that appeared once came back in later runs.

Sample variance across four runs of the same prompt in the pizza example shows how this plays out. The question was the best pizza place in Manhattan for a tourist with time to try one slice. Across four ChatGPT responses, Johns of Bleecker Street and Joe’s Pizza Broadway appeared in every reply, while Prince Street Pizza, L’Industrie Pizzeria West Village, Lombardi’s, and NY Pizza Suprema each surfaced only once. Average similarity between runs was 43%, and average persistence was 50%.

How much do AI answers actually change?

Across repeated runs of the same prompt, the overlap between result sets was low: 23% for ChatGPT, 21% for Google AI Mode, and 33% for Google AI Overviews. Persistence was higher than similarity, with a business that appeared once returning in about half of subsequent attempts on ChatGPT and AI Mode and 58% on AI Overviews. Still, 71 to 80% of businesses showed up in half the attempts or fewer, so inconsistent inclusion is the rule rather than the exception.

The Google AI Overviews surface was the most consistent of the three, in part because it grounds its answers in traditional search results and in part because it returns the shortest lists. AI Mode saw the largest changes between runs, with ChatGPT close behind.

How many businesses does each platform name?

ChatGPT named the most businesses per response, averaging 4.1 with a range of 1 to 7 and a median of 4. Google AI Mode averaged 3.5, with a range of 0 to 6 and a median of 4. Google AI Overviews was the most limited, averaging 2.5 with a range of 0 to 4 and a median of 3. Empty responses were not rare: 10% on ChatGPT and 11% on each of the two Google surfaces.

Because responses can contain different counts and mixes of businesses, traditional similarity scores understate how much the answers actually move.

Does Google Maps ranking translate into AI visibility?

Not directly. Google Maps mentioned a tracked business in 66% of searches, while AI surfaces mentioned the same business only 32 to 38% of the time: 38% on AI Overviews, 33% on ChatGPT, and 32% on AI Mode. Maps was also far steadier, with 99% agreement across attempts compared to 91 to 96% for the AI platforms.

The platforms disagree with each other, too. Google AI Mode and Google AI Overviews shared only 29% of named businesses on average, despite both leaning heavily on Google Business Profile as a source. ChatGPT overlapped only about 20% with AI Overviews and 19% with AI Mode, a reflection of ChatGPT’s stronger reliance on Yelp and Bing rather than Google Business Profile.

Google Business Profile made up 28.5% of all citations across the three platforms. All three still pull from Yelp, and Google’s surfaces pull from Facebook as well, which helps explain why ChatGPT’s lists look different from Google’s even when the underlying prompt is identical.

Why AI answers shift: fan-out queries

AI search engines answer a prompt by breaking it into several related sub-questions, searching for each one separately, and blending the results. The team called these fan-out queries. A prompt such as best lawyer in downtown LA for family law might trigger separate searches for best family law attorney Downtown Los Angeles, divorce custody reviews Los Angeles, and certified family law specialist downtown Los Angeles. Each sub-search can return its own candidate set, and the AI stitches them together. That extra layer of retrieval is one of the main reasons AI answers fluctuate while a single keyword-driven Local Pack stays more stable.

Does moving across town change AI results?

Yes, and ChatGPT is the most volatile. Overlap between points in the same city was 36.2% on ChatGPT, 46.9% on AI Mode, and 47.4% on AI Overviews. Only 3.5% of ChatGPT results were always present across the map grid, compared to 8.8% for AI Mode and 25.0% for AI Overviews.

Proximity still matters across all three platforms. Between 52.8 and 59.3% of picks fell within 5 km of the searcher, and 71.7 to 78.2% within 10 km. Beyond that radius the platforms diverge sharply: ChatGPT’s circle of influence, the radius covering 90% of results, stretched to 287 km, compared to 64 km for AI Mode and 38 km for AI Overviews. Google Business Profile, location pages, and consistent NAP across listings all feed into that proximity signal.

What this means for local businesses

Showing up in roughly half of AI runs for a given prompt is a workable baseline, but appearing in only one of ten runs is a different situation. Volatility is real, but it is not uniform, and the businesses that recur more often tend to be the ones actively building the signals AI surfaces pull from. Tracking AI presence over time and noting which competitors appear alongside a business gives a clearer picture than spot checks.

Practical moves that hold up across AI and traditional search:

• Keep name, address, and phone number consistent across every listing and mention.
• Make the business address easy to find on the website.
• Build out clear location pages with the information an AI agent would need to recommend a business.
• Keep accumulating reviews and listings on the directories the AI surfaces rely on, including Yelp and Facebook as well as Google Business Profile.
• Tune the Google Business Profile category so it matches the queries a business wants to win.

Rank tracking across a service area, mapped town by town, is what a geo grid report does, and it shows exactly where a business is invisible rather than giving a single city-wide average. BizScoreAI scores how visible a business is to AI search and shows what its listing looks like to the engines people ask.

FAQ

How often does AI mention the same local business across repeated searches?

In the study, a business that showed up once returned in about 50% of follow-up runs on ChatGPT and Google AI Mode, and 58% on Google AI Overviews. Google Maps mentioned the same business 66% of the time and held 99% agreement across attempts.

Why do AI search results change so much for local businesses?

AI engines break a single prompt into several sub-questions, search for each one separately, and blend the answers, a process the researchers called fan-out queries. Because that extra layer of retrieval can pull different candidates each time, the same prompt returns a different mix of businesses on most runs.

Does ranking well on Google Maps guarantee AI visibility?

No. Google Maps mentioned a tracked business 66% of the time, while AI surfaces mentioned the same business 32 to 38% of the time. The Local Pack uses a different algorithm than the AI surfaces, and strong Maps performance does not automatically carry over.

BizScoreAI

BizScoreAI scores how visible a business is to AI search and shows what its listing looks like to the engines people ask. Open the AI visibility scan.