Free AI Visibility Prompt Generator

Free tool: generate around 50 buyer questions to test in ChatGPT, Gemini and Perplexity, grouped by intent, with a tracking sheet for brand mentions and cited sources.

Enter what you sell, where, and your brand name. The tool writes a set of around 50 buyer questions to test in ChatGPT, Gemini, Perplexity and Copilot, grouped by intent, with a tracking sheet to record whether you were named and which sources were cited.

Free tool AI Visibility Prompt Set Generator Runs in your browser. Nothing you paste is sent anywhere.

Why you need a fixed prompt set

The only way to know whether AI assistants recommend your business is to ask them and write down what they say. Most people do this once, with two or three questions they thought of on the spot, and learn nothing, because assistants give different answers on different days and a single test cannot separate a real change from noise.

A fixed set solves that. You ask the same questions, the same way, every month. Over three or four rounds a pattern appears: which kinds of question you show up for, which you never do, and which competitors hold the answers you want.

The seven kinds of prompt

Category Example What it shows
Discovery best CNC machining in Pune Whether you are recommended to someone who has not heard of you. The one that matters most
How to choose what should I look for in a CNC machining company Whose advice the assistant repeats. Being the source here shapes the buyer’s criteria
Price how much does CNC machining cost in Pune Whether your published prices are being used
Trust is Acme Precision reliable What the assistant says when a buyer checks up on you
Brand what does Acme Precision do Whether the assistant knows you exist and gets the facts right
Problem how long does CNC machining take Whether your educational content is cited
Local CNC machining in Pune with good reviews How your directory listings and reviews are being read

If you add competitors, the tool also writes comparison prompts for each one.

How to run a test properly

  1. Download the tracking sheet and add any questions your sales team hears often. Real buyer wording beats anything generated.
  2. Use a fresh chat for every prompt. An assistant that has already discussed your brand will favour it in the next answer.
  3. Sign out, or use a private window, so your own history does not colour the result.
  4. Record four things: were you named, was your domain cited as a source, which competitors were named, and which sources were cited.
  5. Repeat with the same set every month. Change the set only by adding to it.

The “sources cited” column is the most useful one. It tells you which websites the assistant trusts for your category. If the same three directories or publications appear every time, being listed or mentioned there is your next job.

Reading the first round

Start with the Brand prompts. If an assistant cannot say correctly what your company does and where it is, nothing else will work yet, and the fix is basic: clear pages, consistent listings, structured data. If Brand is fine but Discovery is empty, the assistant knows you and does not rank you among the recommendations, which is a question of reputation and third party mentions rather than of your own website.

What this tool cannot tell you

It writes questions. It does not run them or score the answers, and it has no data on how often real buyers ask each one. The prompts are built from templates, so read them through and delete any that do not fit your trade. Paid tracking tools automate the running and counting, and are worth considering once you have done two manual rounds and know what you want to measure.

Frequently asked questions

How many prompts do I need?

Thirty to fifty is enough to see a pattern and few enough to finish in an afternoon. Coverage across the categories matters more than volume.

Which assistants should I test?

ChatGPT, Gemini and Perplexity cover most use. Add Copilot if your buyers work in companies that run Microsoft 365.

The answers change every time. Is the test worthless?

No, but it means one answer is an anecdote. Count how often you appear across the whole set and compare that figure month to month.

Related