Rarely: when 38 businesses checked their websites with our free check between 20 and 30 October 2026, ChatGPT named the business in 17% of its answers to questions its customers would ask, Perplexity in 23%, and 17 of the 38 were not named in a single answer. Local businesses, whose questions were about their town, did much better than businesses that sell online or to other businesses, and in 400 answers from our detailed reports, 78% of the pages the assistants cited were competing businesses' own websites.
This matters because customers already ask assistants for recommendations. In BrightLocal's Local Consumer Review Survey 2026, published on 11 February 2026 and based on 1,002 US adults, use of ChatGPT and other generative AI tools for local recommendations rose from 6% the year before to 45%. Pew Research Center's Americans and AI 2026, from a survey of 5,119 US adults on 17 to 23 February 2026, found that 44% use ChatGPT, up from 34% a year earlier, and that about four in ten use chatbots for information searching.
Below is what we measured, how we measured it, and what the numbers can and cannot tell you.
How did we measure how often assistants name a business?
The numbers come from two sets of checks that businesses ran on their own websites with ItGrows in late October 2026. Every business is anonymised, and none is described here.
The free check, 20 to 30 October 2026. 38 businesses checked their websites. For each one, five questions its customers would ask were generated from its own website. Each question was put once to ChatGPT, with web search, and once to Perplexity. That gave 365 answers in all. Perplexity was unavailable for part of 30 October, so it answered 175 questions and ChatGPT 190. For every answer, we noted whether it named the business.
The detailed report, 27 to 30 October 2026. Five businesses got the detailed free report: 20 customer questions each, asked once to ChatGPT, Claude, Gemini and Perplexity, 400 answers in all. Here we also recorded the web pages the answers cited, 478 of them.
The questions are ones a customer would ask, so the test is whether an assistant brings the business up on its own.
How often were the businesses named?
In the free check, the typical business was barely there. 17 of the 38 businesses (45%) were not named in a single answer. The median business was named in 10% of its answers and the average business in 19%. The average sits higher because a few businesses were named often: only 6 of the 38 were named in half of their answers or more.
The detailed report, with four times as many questions per business and four assistants, points the same way. The five businesses were named in 89 of 400 answers (22%), from 9 to 26 of 80 per business. Every one of the five was named somewhere, yet even the most visible was missing from about two thirds of its answers.
| Assistant | Free check, 20-30 October 2026 (38 businesses, 5 questions each) | Detailed report, 27-30 October 2026 (5 businesses, 20 questions each) |
|---|---|---|
| ChatGPT | named the business in 17% of answers (32 of 190) | 21% of answers |
| Perplexity | 23% of answers (41 of 175) | 22% of answers |
| Claude | not asked | 25% of answers |
| Gemini | not asked | 21% of answers |
The differences between assistants are small. Perplexity named businesses a little more often than ChatGPT in the free check; in the detailed report all four sat between 21% and 25%. With one ask per question, a gap of a few points could be chance, as the limits section below explains. The bigger story is not which assistant, but which business and which question.
Do local businesses get recommended more often than online and B2B businesses?
Yes, by a wide margin in our data. Of the 38 businesses in the free check, 17 were local businesses, and their questions were about their town. Only 2 of the 17 were never named, and on average they were named in 25% of their answers.
The other 21 sell online or to other businesses. 15 of them (71%) were never named, and they were named in 14% of their answers on average.
Our reading of this, which the data does not prove: a question about a service in one town has a short list of possible answers, and a business that serves that town is more likely to be on it. A question about software, a product sold online or a B2B service is answered from a far larger field, where many companies compete for the same few names in an answer.
If you sell online or to other businesses, what changes for B2B companies is the place to start. If you serve one area, how local businesses get recommended by AI assistants goes further into the local case.
Does it help when the customer's question names your town?
Yes. Answers to questions that named the business's town named the business in 35% of cases (31 of 89). Answers to all the other questions named it in 15% (42 of 276).
So the way a customer phrases the question makes a large difference. A question with the town in it narrows the field to businesses that serve that place. A question without a place leaves the assistant free to name any business anywhere, or none.
For an owner, that means two things. First, the questions with your town in them are where your chances are best, so check whether you are named there. Second, do not stop there: most of the questions in our check did not name a town (276 of 365), and on those the business was named less than half as often.
Which websites do the assistants cite instead of yours?
In the detailed reports, the answers cited 478 web pages:
- 373 (78%) were the websites of competing businesses;
- 51 (11%) were official sources, such as health services and regulators;
- 27 (6%) were directories;
- 17 (4%) were reference sites.
For an owner, this is the most useful number in the study. When the assistants answered customer questions, they were not mainly quoting directories or reference sites. They were reading other businesses' own websites.
A likely reason, which our counts cannot prove: when a competitor's page answers the customer's question directly, the assistant has something to read and cite, and the business behind that page sits in the answer's sources. If your own site does not answer the same question, the assistant has nothing of yours to work from, however well known you are locally.
What should a small business owner do with these numbers?
Get your own number first. The averages hide a wide spread: 17 businesses in the free check were never named, while 6 were named in half their answers or more. You could be anywhere on that range. Our free check generates five questions your customers would ask from your homepage and shows answers from ChatGPT and Perplexity; two answers show in full right away, and all five open once you create an account. The detailed report, 20 questions each put once to all four assistants, is free once you have an account; here is a real one.
Answer your customers' questions on your own website. Since competitors' own websites made up 78% of the cited pages, a page that clearly answers a real customer question is what an assistant can use: what you do, where, for whom, what it involves and how to choose. Cover the questions with your town in them and the general ones without it.
Make sure assistants can read your site. If your hosting or security settings turn AI crawlers away, nothing on the site can be read; check whether your website is blocking them.
Measure the same questions over time. One check is a snapshot. Compare like with like: the same questions, the same assistants, the same number of asks. How to measure AI visibility over time sets out how.
If you would rather have the work done for you, the ItGrows plan does it on your own website and, every two weeks, puts your customers' questions to ChatGPT, Claude, Gemini and Perplexity and counts the answers that name you. It costs $499 a month plus a one-time $499 setup; the setup is free when you pay $1,497 for 3 months or $4,788 for a year ($399 a month). It carries one guarantee: "In 90 days, AI assistants will name you more often than on launch day — or we work for free." Conditions apply.
Method and limits: what can this study tell you?
A small sample. 38 businesses in the free check and 5 in the detailed report. The percentages describe these businesses, not small businesses in general.
One ask per question. Each question went to each assistant once, and assistants rarely give the same answer twice. In SparkToro's study, published on 28 January 2026, 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI Overviews 2,961 times in November and December 2025, and the tools gave the same list of brands less than once in 100 times. A business missing from one of our answers might appear in the next run, and the reverse.
The businesses chose to check themselves. They came to us; they were not drawn at random. Owners who already suspect a problem may be more likely to run a check.
Questions generated from each website. In the free check, the five questions came from each business's own website. They reflect what the website says the business does, not a record of what its real customers type.
Uneven coverage. Perplexity was unavailable for part of 30 October, so it answered 175 free-check questions against ChatGPT's 190. The free check asked only ChatGPT and Perplexity; Claude and Gemini appear only in the five detailed reports.
A snapshot. All answers were collected between 20 and 30 October 2026. Assistants change their models and their search sources, and people's habits are moving fast: in the Reuters Institute's Generative AI and news report 2025, published on 7 October 2025 and covering six countries, 24% of people had used generative AI for getting information in the previous week, up from 11% a year before. Treat our numbers as a baseline to compare against, not a constant.
Frequently Asked Questions
Does ChatGPT recommend small businesses?
It does, but rarely for any one business. In our free check of 38 businesses from 20 to 30 October 2026, ChatGPT with web search named the business in 17% of its answers to questions its own customers would ask (32 of 190). In the five detailed reports, it named the business in 21% of its answers.
Does Perplexity recommend small businesses more often than ChatGPT?
Slightly, in our data. Perplexity named the business in 23% of its free-check answers (41 of 175), against 17% for ChatGPT; in the detailed reports the two were almost level at 22% and 21%. With one ask per question, treat the difference as small.
Why does an AI assistant name my competitors and not me?
In our detailed reports, 78% of the pages the assistants cited (373 of 478) were competing businesses' own websites. If a competitor's site answers the customer's question and yours does not, the assistant has their page to work from and not yours.
How can I find out how often AI assistants name my business?
Run the free check on your homepage: it puts five questions your customers would ask to ChatGPT and Perplexity, shows two answers in full right away and all five once you create an account. For the fuller picture, the detailed report with 20 questions across ChatGPT, Claude, Gemini and Perplexity is free once you create an account.