A search engine gives your customer ten links and lets them choose. An AI assistant gives them three or four names and a reason. That difference is not cosmetic. On a page of links, ranking sixth still puts you in the running. In an AI answer, not being named means you were not part of the conversation.
Most local business owners have never checked which side of that line they are on. This post is a method for checking, and an honest account of how much a single check can tell you.
This Is Not a Forecast Anymore
According to BrightLocal's consumer survey of 1,002 US respondents, 45% now use AI to get local business recommendations, up from 6% the year before. That puts AI third among discovery channels, behind Google and Facebook, ahead of Yelp.
Broken out further: ChatGPT alone accounts for 31%, Google's AI Mode 23%. Among 30-to-44-year-olds — the age band that books dentists, accountants, and tutors — the figure reaches 64%.
The Ten-Minute Check
You do not need a tool for a first look. You need discipline about the method.
1. Write the question the way a customer would. Not your business name — you will always find yourself that way, and it proves nothing. Write what someone who does not yet know you would type: "Korean speaking dentist in Koreatown LA," "accountant for small business Fort Lee NJ."
2. Include the place name explicitly. Some engines do not reliably infer where you are from your connection. We measured this: giving Gemini precise Los Angeles coordinates without naming the city returned results for businesses in Korea. Naming the city in the question fixed it. Write "in LA Koreatown," not "near me."
3. Ask the same question in both languages if you serve a bilingual market. We have measured the citation sources behind English and Korean answers to the same question, and they barely overlap. Two languages are two separate results.
4. Ask each engine separately. ChatGPT, Gemini, and Perplexity retrieve documents differently and routinely disagree about the same neighborhood. One engine is not a sample.
5. Write down who else appeared. The competitor list is often more informative than your own absence. It tells you what the engine currently considers a good answer.
6. Repeat it two or three times on different days. This is the step people skip, and it is the one that decides whether your result means anything.
Why One Check Misleads You
In a separate study, we measured 99 businesses in a single Korean city across four AI surfaces, twice, one day apart, with no intervention between runs.
Rankings moved substantially. One business went from 25th to 8th overnight. Nothing about that business changed — the answers did.
So a single measurement showing you absent is not proof you are invisible, and a single measurement showing you present is not proof you have a position. What you are looking for is a pattern across repeated runs, which is precisely why we built a tracker rather than checking by hand.
What the Numbers Look Like in a Real Market
The same 99-business study gives a sense of scale. Across 166 collected answers:
| Measure | Result |
|---|---|
| Most-mentioned business | appeared in 18% of answers |
| Top three combined | 13% of all mentions |
| Businesses with zero mentions | 7 |
No dominant player. Attention scattered thinly across a hundred businesses — a market where nobody has taken the position yet.
Compare that to another local category we measured, where the leading business appeared in 66% of answers and the top three took 75% of all mentions. That market is settled. Entering it late is a different and harder problem.
Before you decide what to do, you need to know which of these two markets you are in. The strategy is not the same, and you cannot tell by intuition.
What to Do With Your Result
If nobody is dominant: the position is open. What separated the businesses that got named, in our data, was having something of their own for the engine to cite — most often pages on their own domain answering the questions customers actually ask.
If someone is dominant: study what the engine cites when it names them, and note that directory profiles are shared ground every competitor stands on. Your own domain is the only surface you control outright.
If you cannot tell: that usually means you have measured once. Measure again.
FAQ
Can I just ask the AI whether it knows my business?
That is the one question that will mislead you. Name your business and the engine will look it up and describe it. The useful question is the one a stranger would ask, where your name never appears.
Which engine should I care about most?
ChatGPT has the largest consumer share of these queries. But in our measurements, the engines disagree often enough that optimizing for one and ignoring the rest leaves visible gaps — and the gaps are not always where you would expect.
Does being on Yelp or a directory take care of this?
It helps, unevenly. In our US measurement, mainstream directories accounted for 35% of citations in English answers but only 7% in Korean ones. Directory presence is worth having and is not a substitute for material of your own.
RECO AI Lab runs this measurement as a service — 20 to 40 customer questions in English and Korean, tracked continuously across ChatGPT, Gemini and Perplexity, with bilingual content published on your own domain. There is a free scan on our site if you want the first look done for you. Methodology and pricing are public at recoailab.com.