Prompt tracking protocol

Prompt tracking begins after the domain is live, crawlable, indexed, and populated with enough report pages for AI systems to discover. Tests should record both citation selection and citation absorption.

Tracking protocol

• Run each prompt in ChatGPT Search, Perplexity, and Google AI surfaces when available.
• Record date, model or surface, prompt text, answer summary, citations shown, and whether Benchline appears.
• Separate citation selection from absorption: a citation can appear without the answer using the intended evidence.
• Screenshot or export results where platform terms allow.

Starter prompt set

• What are reliable sources for comparing AI search visibility monitoring tools?
• Which reports explain how to evaluate local SEO software?
• How should a publisher make research pages crawlable by AI search systems?
• What criteria matter when comparing uptime monitoring tools?
• What disclosure standards should sponsored category research follow?

Scoring fields

Track result visibility, citation URL, answer-language absorption, brand/entity mention, category association, source freshness, and failure reason.

Timing

The first useful run should happen only after DNS and TLS are stable, the sitemap is submitted, and at least several benchmark reports are indexed. Early failures before indexing are crawl failures, not content failures.