Build a prompt set you can test repeatedly
Fill in what you know, choose a prompt type, then copy the prompts into your test log. The page does not send your entries anywhere. For the rationale and a text-only version, read 18 AI citation test prompt examples.
Category options
What are the best your product or service category options for your buyer in your market?
Provider discovery
Which companies help your buyer with your product or service category?
Buying criteria
What should your buyer look for when choosing a your product or service category provider?
Brand comparison
Compare your brand with other your product or service category options for your buyer.
Alternatives
What are alternatives to your brand for your product or service category?
Best-fit shortlist
Which your product or service category providers are best for your buyer, and why?
Entity definition
What is your brand?
Offer retrieval
Does your brand offer your product or service category for your buyer?
Source retrieval
Which pages on example.com explain your brand's your product or service category work?
Problem solution
How can your buyer solve your product or service category?
Improvement path
What is the best way for your buyer to improve your product or service category?
Buyer questions
What questions should your buyer ask before buying your product or service category?
Market provider
Who provides your product or service category in your market?
Availability
Is your brand available for your market customers?
Hiring checks
What should your market buyers check before hiring a your product or service category provider?
Evidence check
What evidence supports your brand as a your product or service category provider?
Tradeoffs
What limitations or tradeoffs should I know about your brand?
Source support
Which sources support recommendations for your product or service category options?
Which prompt types should I test first?
| Prompt type | What it checks | What to watch |
|---|---|---|
| Brand-free discovery | Whether the brand appears when the user has not named it. | Do not call one answer a ranking. |
| Comparison | Whether the brand is described beside alternatives. | Separate mention, citation, and accuracy. |
| Known-brand retrieval | Whether the answer can identify the brand and find its source pages. | Check whether cited pages actually support the statement. |
| Problem or need | Whether the page maps to the job a buyer is asking about. | Freeze the prompt before looking at results. |
| Local or service fit | Whether the answer handles service area, market, and buyer fit. | Record market, language, account, and visible mode. |
| Evidence and credibility | Whether claims, tradeoffs, proof, and sources travel with the answer. | Do not count unsupported praise as evidence. |
How to use the prompts responsibly
- Freeze the prompt set before checking any answer.
- Run the same prompt more than once before calculating a rate.
- Record the product, date, mode, market, account state, and page version.
- Log mentions, visible own-domain citations, cited third-party sources, and incorrect statements separately.
- Use the AI citation test workbook before using the AI citation rate calculator.
If you need the full prompt list and customization guidance, read 18 AI citation test prompt examples. For the measurement method, read how to check whether AI systems cite your website. If you want to interpret a result, start with why one ChatGPT result is not a ranking.
What these prompts cannot prove
A prompt set is not a benchmark, forecast, causal proof, or permanent visibility claim. It is a measurement aid. AI answers can change by product, date, account state, visible mode, market, language, retrieval behavior, and source availability.
Audit the page before testing prompts.
The prompt set can reveal citation behavior. It cannot fix a page that is unclear, unsupported, or hard to quote.
Audit a page free