Most business owners have never asked ChatGPT whether their business exists. That sounds like a strange thing to overlook, but it's genuinely common — and it's the first thing worth fixing.
Here is how to run the test properly.
Start with category queries, not brand queries
The mistake most people make is searching for their own name. Typing "tell me about [Your Business Name]" into ChatGPT tells you almost nothing useful. If you're well-known enough to have a Wikipedia entry, you'll show up. If you're not, you probably won't. That's not the test.
The test is: what happens when someone who has never heard of you asks for what you sell?
Use queries like:
- "What are the best [your category] options for [your typical customer]?"
- "Can you recommend a good [your service] in [your city]?"
- "I'm looking for a [your product/service] — what do you suggest?"
- "What should I look for in a [your category] provider?"
These are the actual queries your potential customers are asking. If you appear in the answers, good. If you don't, that is your baseline.
Run it across more than one model
ChatGPT, Claude, Perplexity, and Gemini handle these queries differently. A brand that shows up consistently in Perplexity may be invisible in Gemini. Models that use live web retrieval (Perplexity, ChatGPT with browsing enabled) will give you more recent results than models reasoning from training data alone.
Run the same query across at least three of these. Note the differences.
What to look for in the responses
Whether your brand appears at all. Being mentioned anywhere is meaningfully different from not being mentioned.
Where in the response you appear. A brand named first is getting a different recommendation than a brand listed fifth. Order reflects the model's implicit ranking.
Whether your competitors appear instead. If your direct competitors are showing up and you aren't, you now know the gap exists and roughly how large it is.
What the model says when it mentions you. Is the description accurate? Is it citing a URL? Is there any qualifier that signals uncertainty?
The limitations of manual testing
Manual testing is useful for orientation, but it has real limits.
Responses vary. Ask the same question twice and you may get meaningfully different answers. A single test doesn't give you signal — it gives you an anecdote.
It's also time-consuming to do this properly across multiple models, multiple query variations, and multiple competitors. Most people run one query, get one result, draw a conclusion, and move on.
And you can't track trends. The manual test tells you where you stand today. Without a repeating benchmark, you have no way to know if things are improving or getting worse.
A more reliable way to measure it
The free audit at SearchVisible runs a structured set of queries across ChatGPT, Claude, Perplexity, and Gemini and scores the results using the V-Score methodology — not just whether you appear, but where, and how that compares to your competitors.
It takes about two minutes and gives you a repeatable baseline rather than a one-off data point.
Run your free audit and see where your brand stands across the four major AI models.