BeAIWho do AI assistants recommend for local services? We asked 960 times in 40 US cities
Short answer: In our test of 40 US cities (960 answers), ChatGPT and Perplexity named no business in common 63% of the time. So being recommended by one AI doesn't mean you're recommended by the others, and our report checks all five: ChatGPT, Gemini, Grok, Claude and Perplexity.
Customers now ask AI assistants questions like “Who is the best plumber in my city?” To see what comes back, we asked two of them, ChatGPT and Perplexity, that question for four trades in 40 US cities, three times each. Here is what we found, as patterns only. We don't rank or rate any business on this page.
▶ Watch the 2-minute video summary on YouTube (one example from the study: plumbers in Columbus, Ohio).
▶ Watch: dentists in Dallas, Texas (another example from the study).
This study tested ChatGPT and Perplexity only. Our full $49 report checks five AIs (ChatGPT, Gemini, Grok, Claude and Perplexity), because they disagree this much.
1. The AIs name only a few businesses
Asked for “the best”, both AIs almost always said there is no single best (96% of ChatGPT answers and 92% of Perplexity answers said so in some form), and then named a short list anyway. ChatGPT named 4.3 businesses per answer on average. Perplexity named 2.0, and in 25% of its answers it named just one.
Average number of businesses named per answer, by trade.
So in a city with hundreds of plumbers or dentists, a customer who asks AI sees two to five names.
2. The two AIs rarely agree
For the same question in the same city, ChatGPT and Perplexity gave the same top pick only 9% of the time. In 63% of cases, their lists had no business in common at all. Even when we pooled all three tries from each AI, 24% of city-and-trade pairs still had no overlap.
How often ChatGPT's and Perplexity's answers named at least one of the same businesses, by trade.
Dentists had the least overlap. Part of the reason: ChatGPT often named individual dentists (70% of its dentist answers named a “Dr.”), while Perplexity usually named practices (18%).
3. Ask again and the answer can change
We asked every question three times. ChatGPT gave the same top pick all three times for 6% of questions, and Perplexity for 25%. Across its three answers to one question, ChatGPT named 9.9 different businesses on average, and Perplexity 4.1.
One check of one AI is a snapshot, not a verdict.
4. Where the AIs got their information
We sorted every website each AI cited or pulled up into types. They leaned heavily on “best of” list sites (pages that rank local businesses), BBB, review sites, directories and the businesses' own websites.
Share of answers that cited at least one site of each type. Perplexity returns more sources per answer (5.9 sites on average vs 1.9 for ChatGPT), so its bars run longer. *Business websites are estimated from the web address.
Yelp showed up in 2% of ChatGPT answers and 21% of Perplexity answers. BBB appeared in 17% and 56%.
5. Big cities vs. mid-size cities
We split the 40 cities into 20 of the largest US cities and 20 mid-size cities (roughly 150,000 to 500,000 people), with the same four trades and the same question.
| 20 big cities | 20 mid-size cities | |
|---|---|---|
| Businesses named per answer, ChatGPT | 4.3 | 4.3 |
| Businesses named per answer, Perplexity | 1.9 | 2.2 |
| Both AIs gave the same top pick | 9% | 10% |
| Both AIs named at least one of the same businesses | 38% | 35% |
| Same top pick all 3 times, ChatGPT | 4% | 8% |
| Same top pick all 3 times, Perplexity | 28% | 21% |
| Different businesses named across 6 answers (avg.) | 12.4 | 12.8 |
| One business named in 3+ of the 6 answers | 84% | 85% |
| Answers naming a national chain or franchise | 5% | 5% |
The patterns held in both groups. The clearest difference: in big cities, Perplexity named just one business in 34% of answers, compared with 16% in mid-size cities. Sources shifted too. BBB showed up in 42% of mid-size city answers vs 31% in big cities, and licensing and government sites in 19% of big-city answers vs 7%. Chains and franchises were rare in both, so in our test the names were mostly local businesses.
“National chain or franchise” is a short list of well-known brand names we checked for, so treat it as a rough count.
What this means for a local business
Because each AI reads different websites and names only a few businesses, being well known in your town doesn't guarantee AI will name you. One small step that helps with every AI: make sure your business name, phone number and address match everywhere you're listed, including BBB and the directories in your trade. The rest depends on your business, and which sites each AI is reading for your town.
How we tested
- Date: October 25, 2026.
- Question: “Who is the best [plumber / dentist / HVAC company / roofer] in [city, state]?”, asked word for word, with the location set to that city.
- Cities: 20 of the largest US cities and 20 mid-size cities in 27 states. Mid-size means roughly 150,000 to 500,000 people; population figures are approximate.
- AIs: ChatGPT (OpenAI's API with web search) and Perplexity (Sonar, with web search). We did not test Gemini, Grok or Claude for this study.
- Answers: 160 questions, each asked 3 times of each AI: 960 answers came back usable.
- Counting: we pulled out the businesses each answer named, left out directories, agencies and websites, and matched names across answers allowing for small spelling differences, so name matching, and the counts built on it, are approximate. Website types are grouped from each answer's cited web addresses.
- Limits: all answers were collected on one day (October 25, 2026). AI answers change from day to day and person to person. These are API answers, which can differ from what you see in the apps. This is a snapshot of patterns, not a ranking of any business, and it says nothing about the quality of anyone's work.
Common questions
Do ChatGPT and Perplexity recommend the same local businesses?
Not often, in our test. For the same question in the same city, they gave the same top pick 9% of the time, and 63% of the time they had no business in common.
How many businesses does AI name when asked for the best one?
In our test, ChatGPT named 4.3 businesses per answer on average and Perplexity named 2.0.
Is this a ranking of the best businesses?
No. It shows patterns in what two AI assistants said on one day. We don't rank, rate or name any business on this page.
Related guides
- What is AI search, and why it matters
- Get your business recommended by ChatGPT
- Why AI recommends your competitor, not you
- Study results for plumbers
- How drain cleaners get recommended by AI: a 20-answer test
- How window replacement companies get recommended by AI: a 20-answer test
- How painters get recommended by AI: a 20-answer test
- How tree services get recommended by AI: a 20-answer test
- How flooring companies get recommended by AI: a 20-answer test
- How electricians get recommended by AI: a 20-answer test
- How pest control companies get recommended by AI: a 20-answer test
- How concrete contractors get recommended by AI: a 20-answer test
- How insurance agents get recommended by AI: a 150-answer test
- Study results for dentists
- How HVAC companies get recommended by AI assistants
- How roofers get recommended by AI assistants
- All guides
Does AI name your business?
Our $49 report asks ChatGPT, Gemini, Grok, Claude and Perplexity 15 customer questions about your business in your town and shows what to fix first.
Get the $49 reportSee a sample report (insurance agency example)See the sample (insurance example)Yours is custom built from your own test results.