We asked ChatGPT, Gemini and Google AI for the best local businesses in 8 US cities. 84% of 2,926 were never named

When someone asks ChatGPT for "the best dentist in Austin", a handful of practices get named and everyone else is invisible. We wanted to know how many get named, whether the engines agree, and what the named businesses have in common. So in November 2026 we took 4 local categories in 8 US metros, pulled every business Google Maps returned for them, asked ChatGPT, Gemini and Google AI Overviews the questions a customer would ask, and checked each answer against the full list. The per-business results, without names, are in a CSV you can download.

The short version

• 84% of the 2,926 businesses were not named in any answer. Only 472 (16.1%) were named at least once across 178 AI answers.
• The engines rarely agree. Of the businesses named in markets where all three engines answered, 71.5% were named by only one engine and just 4.5% by all three.
• Reviews help but do not decide. Businesses with 1,000+ Google reviews were named 34.3% of the time, against 9.4% for those under 50. Yet 65% of the five most-reviewed businesses in each market were never named.
• Star rating tells you almost nothing here. The median rating was 4.9 for named and unnamed businesses alike, because 86% of the businesses we checked are rated 4.8 or higher.
• Structured data is the clearest website signal. Sites with LocalBusiness-type schema on the homepage were named 17.2% of the time, against 11.7% without it.
• Each category has its own referees. Answers about lawyers leaned on Avvo and Super Lawyers, agencies on Clutch, dentists on Google reviews and ranking sites, med spas on Yelp.

Most local businesses are invisible to AI

A typical answer named about four local businesses. With around 120 dentists in each metro's Maps results, that leaves most practices out of every answer, however good they are. Dentists fared worst, partly because there are simply more of them per city. The share named ranged from 12.3% of businesses in Phoenix to 21.3% in Nashville.

These rates are higher than the 1.2% ChatGPT recommendation rate in SOCi's 2026 Local Visibility Index, because we count a business as named if any of three engines named it in either of two questions, and because SOCi measured locations of multi-location brands rather than mostly independent businesses. The direction is the same: AI answers pick a few and ignore the rest.

The engines rarely agree

We got a Google AI Overview for 50 of the 64 questions; the other 14 returned none or failed to load. More striking is how little the lists overlap. Of the businesses named in markets where all three engines answered, 71.5% appeared in only one engine's answers and only 4.5% were named by all three. Being recommended by ChatGPT says little about whether Gemini or Google will recommend you, so checking one engine is not enough.

Reviews matter, up to a point

Review volume is the strongest signal we found. The median named business had 322 reviews, against 146 for businesses that were not named, and businesses past 500 reviews were named about three times as often as those under 50. Within each market, the median named business sat in the top 27% by review count.

But volume is not a guarantee. 65% of the five most-reviewed businesses in each market were never named, and 72 businesses with fewer than 50 reviews were. Review count gets you into the conversation; it does not decide it. Star rating barely mattered in our data, because nearly everyone is rated 4.8 or higher. Only 2 of the 472 named businesses were rated under 4 stars.

What the websites had in common

We fetched the homepage and robots.txt of every business with a website. 2,085 sites returned a page we could read; the rest blocked automated requests or failed to load.

Businesses whose homepage declares what they are, with LocalBusiness-type structured data, were named about 1.5 times as often as those without it. FAQ markup and a phone number on the page made no clear difference. Twelve sites blocked the AI search crawlers in robots.txt; none of them was named. The robots.txt groups are small, so treat those last two rows as a warning sign, not a measurement.

These are correlations. Practices that bother with schema may also have better listings, more reviews and more press. We cover what the wider evidence says in does schema markup help you get cited by AI?.

Who the engines quote as referees

AI answers often justify a pick by pointing at a directory or ranking site. For a law firm, a complete Avvo and Super Lawyers profile is part of AI visibility; for an agency, it is Clutch. Google's own reviews and Business Profile came up in every category, which matches BrightLocal's finding that profiles are a large share of the sources behind local AI answers.

What to do with this

1. Check all three engines, with the questions customers actually ask. A single ChatGPT result tells you little about Gemini or Google.
2. Keep earning reviews. It is the strongest signal in our data, and the gap opens up past a few hundred.
3. Add LocalBusiness schema of the right type, with your name, address, phone and hours, to the homepage.
4. Complete your profiles on the directories your category's answers cite: Avvo and Super Lawyers for lawyers, Clutch for agencies, Yelp for med spas.
5. Make sure robots.txt lets the AI search crawlers in. See which AI crawlers to allow.

You can run the same check on your own business for free: Cited Lens asks ChatGPT, Gemini and Google AI Overviews your customers' questions, checks your Google Business Profile and your website, and shows who gets named instead.

Method and limitations

• Markets: Austin, Charlotte, Denver, Nashville, Phoenix, Raleigh, San Diego and Tampa, each with dentists, med spas, personal injury lawyers and SEO agencies (32 markets).
• Businesses: Google Places text searches for each category within 30 km of the city centre, November 2026. We kept the 2,926 operational businesses returned. That is the set a customer would see in Maps, not every business in the city.
• Questions: two per market, phrased the way a customer would ask, for example "Who is the best dentist in Austin, TX?" and "Can you recommend a highly rated cosmetic dentist in Austin, TX?". Each was asked once in ChatGPT, Gemini and Google (AI Overview), between 9 and 13 November 2026. 178 answers came back with text.
• Matching: a business counts as named only if the answer contains its own domain or its distinctive name. Names that read as ordinary prose ("Austin Family Dentistry") need an exact match. This is strict, so it undercounts rather than overcounts.
• One run per question. AI answers vary between runs, so individual businesses can move in and out; the patterns across 2,926 businesses are what we report.
• Website signals come from the homepage and robots.txt only. 724 sites returned HTTP 403 to our fetcher, usually bot protection, and are excluded from that section.
• Everything here is correlation. We did not change any website and watch the result.

Cite or reuse this study

You may quote these numbers and reuse the tables and the CSV with a link to this page. Suggested citation: Cited Lens, "We asked ChatGPT, Gemini and Google AI for the best local businesses in 8 US cities", November 2026, https://citedlens.com/blog/local-business-ai-recommendations-study. Journalists who want the market-level data can write to hello@citedlens.com.

Frequently asked questions