Skip to the content

Research

Who AI assistants recommend, and who they leave out.

We asked ChatGPT, Gemini, Perplexity and Claude the questions people ask when they look for a local business, and kept every answer.

15%

named at least once

Most local businesses never come up.

Of the 12,664 businesses on Google Maps in the markets we checked, 1,954 were named in at least one answer. That is 15%. Our software can miss a name that is spelled in an unusual way, so the real share may be a little higher.

256vs44

median Google reviews

Named businesses tend to have more Google reviews.

The businesses that were named had a median of 256 Google reviews, and those never named had a median of 44. Reviews don’t settle it, though. Among the 10% with the most reviews, 55% were never named.

2%

same first business everywhere

The assistants rarely agree on who comes first.

In 725 cases, we put the same question to several assistants on the same day. They all put the same business first 14 times, which is 2%. Compared in pairs, two assistants named the same business first in 8% of cases.

981

links to Healthgrades

The same few sites keep getting linked.

Healthgrades was linked most often in the answers, 981 times. Next came Yelp (805), Zocdoc (559), the Better Business Bureau (417), WebMD (305), Reddit (304) and Angi (241). These counts cover ChatGPT, Perplexity and Claude. Gemini’s links hide the site behind a Google redirect, so they aren’t counted. Keep in mind that dentists are 35 of the 116 markets.

How we counted

These are 2,388 answers collected between September 2 and October 16, 2026. A market is one place and one line of work, and there are 116 of them. Nearly all are in US cities, 51 in total. Most of them are for dentists, med spas, injury lawyers, HVAC companies, vets, roofers, plumbers or chiropractors. We asked each assistant through its official API, with web search on. Claude has fewer answers (247) because we added it in mid-September. We kept every answer in full, pulled out the business names and matched each one to a Google Maps listing by name and website. The full method is on How it works.