How we check
AI changes its answer. So we ask many times and count.
One look at one answer does not prove much. The same question can name different shops the next time you ask. A count is something you can show a client.
The plan for the monthly count is written. We have not run that count. An Oct 21, 2026 test is posted. It used 12 AI answers per city and trade, not the monthly 30. Read the 7 in 10 report. Two earlier pages still show one ChatGPT answer. See the checked markets.
How to get my business on ChatGPT
What we count
We ask a real customer question. We see if the business is named. We do that many times. The plan uses 30 answers for each AI. The line looks like this: named in ___ out of 30 answers.
The blank is the shape of the monthly count. We have not posted that count. The Oct 21 test is a different, smaller count. We do not turn a count into a rank. We do not blend the AI tools into one score.
How many times
The plan uses six customer questions. We ask each one five times. That is 30 answers for each AI, for each business, each month.
Google’s AI on search results gets 18 answers. We ask it fewer times. Each AI stays in its own count.
Which AI tools
- ChatGPT. The chat tool most people mean when they say AI.
- Gemini. Google’s chat tool.
- Perplexity. An AI tool that shows sources.
- Claude. Another AI chat tool.
- Google AI Overviews. The AI box on Google results. We ask this one fewer times.
The written plan also names one more Google tool. We are not using it yet.
Where the first checks are planned
- Houston, TX
- Phoenix, AZ
- Atlanta, GA
- Tampa, FL
- Denver, CO
- Columbus, OH
- Kansas City, MO
- Sacramento, CA
- Boise, ID
- Albany, NY
We plan to collect answers Oct 21 to Oct 25, 2026. The report page says what we will share.
Rules for posting a number
- No number without a count of asks, the dates, and which AI.
- No rank, and no single score mixed across AI tools.
- A posted count will link to the raw answers. Until then, there is nothing to click.
- We do not give medical, legal, or safety advice. We do not tell you who to hire.
Why one answer is not enough
BrightLocal ran the same local question more than once in a 2026 study. On average, only 20 to 33 percent of the names overlapped from one run to the next. ChatGPT was 23 percent. Google AI Mode was 21 percent. Google’s AI Overviews were 33 percent. That study is theirs, not ours. BrightLocal, 2026 (opens in a new tab)
Questions
Why ask more than once?
AI can change its answer. One try can show your name. The next try can leave it out. A count needs many tries.
When will real numbers show up?
The Oct 21, 2026 test is posted. It used 12 AI answers per city and trade. The monthly count still uses 30 answers per AI, and that count is not posted yet.
Which AI tools are in the plan?
ChatGPT, Gemini, Perplexity, Claude, and Google’s AI on search results. The written plan also names one more Google tool. We are not using that one yet.