How to tell if ChatGPT and Gemini recommend your business
Updated 20 June 2026 · 5 min read
A practical way to check whether AI assistants name your business when buyers ask: which questions to test, how to measure appearance rate, share of voice, and rank, and what the results actually mean.
Key takeaways
- •Test the buying questions your customers ask, not your brand name, and see whether AI names you.
- •Answers vary run to run, so ask each question several times for a directional read.
- •Track three numbers: appearance rate, share of voice, and average rank versus competitors.
- •A "no" usually means the evidence AI needs is missing, not that your product is worse.
- •Doing this continuously catches drift; AgentChoice automates the runs, scoring, and weekly alerts.
How do you check if AI recommends your business?
Ask the AI assistants the buying questions your customers actually ask, several times each, and watch whether they name you, how often, and which competitors they name instead. That is the whole method. The skill is in choosing the right questions, repeating them, and recording the result consistently.
You can do it by hand in ChatGPT, Gemini, Claude, and Perplexity, or automate it. Either way the goal is the same: a repeatable read on whether you show up in the answer where buyers are deciding.
Which questions should you test?
Test the questions a buyer would ask, phrased the way they would phrase them. Do not search your own brand name; that tells you nothing about discovery. Search the category and the intent.
- ›Category and audience: "best CRM for a small team", "project tools for agencies".
- ›Problem-first: "how do I reduce churn in a B2B SaaS".
- ›Comparison: "alternatives to [competitor]", "tools like [competitor]".
- ›Buying modifiers: budget, region, integrations, company size.
Cover the few questions that map to real revenue. Five to ten good questions beat fifty vague ones.
How many times should you ask each question?
More than once. AI answers vary between runs, so a single response is noise. Ask each question several times (five to ten is a reasonable manual minimum) and look at the pattern, not any one answer. If you appear in two of ten runs, that is a 20 percent appearance rate, not a coin flip you happened to lose.
What should you measure?
Reduce each question to three numbers so you can compare over time and against competitors:
- ›Appearance rate: in what share of runs are you named at all.
- ›Share of voice: how much of the named options are you, versus competitors.
- ›Average rank: when named, how high you tend to appear.
Also record who gets named instead of you. That list is your real competitive set in the eyes of AI, and it is often not the competitors you expect.
What do the results mean, and what do they not mean?
A low appearance rate usually means the evidence AI needs to recommend you is missing, buried, or unclear on the pages it can read, not that your product is worse. AI names what it can find and justify.
Be clear on what this measurement is: a model-based, directional snapshot from querying AI models, not a recording of a specific person’s ChatGPT or Perplexity session, and not a prediction of how any individual human will buy. It tells you how AI evaluates the available evidence, which is exactly what you can act on.
How do you do this continuously?
Manually this gets tedious and noisy fast, and your visibility drifts as competitors change their pages and models update. Tracking it over time is what turns a one-off check into a managed metric.
AgentChoice automates it: it runs the queries, computes appearance rate, share of voice, and rank, identifies the competitors named instead, and sends weekly drift alerts when your standing moves. You can run a free check on your own brand to see where you stand today.
Frequently asked
Can I just ask ChatGPT once if it recommends me?
You can, but one answer is unreliable because responses vary between runs. Ask each buying question several times and look at how often you appear, not any single response.
Should I search my own brand name?
No. Searching your brand name only confirms AI knows you exist. To measure discovery, search the category and buying intent, the way a buyer who does not know you yet would ask.
Is this the same as what a real customer sees in ChatGPT?
It is a close, directional proxy. AgentChoice queries AI models directly and samples many runs. It is not a recording of one person’s personal session, which varies with their history and settings, but it reflects how AI weighs the available evidence.
Which assistants should I check?
The ones your buyers use, typically ChatGPT, Gemini, Claude, and Perplexity. Coverage and answers differ between them, so checking more than one gives a fuller picture.