Prompt tracking

Prompt tracking is running a fixed set of the questions your buyers ask in AI engines on a schedule, and recording what each answer says: whether it names your brand, in what position, which competitors appear and which sources it cites. It is the AI search counterpart of rank tracking, with prompts in place of keywords and answers in place of results pages.

Why it matters

Prompts are longer than keywords. Google said in its Q3 2026 earnings remarks that AI Mode queries are on average three times longer than traditional searches, so write prompts as full questions with the details a buyer adds, such as team size, industry or budget. Answers also vary between runs. In a 2026 SparkToro study, AI tools returned the same list of brands less than once in 100 runs, and the author suggested at least 60 to 100 runs per prompt, averaged. A single check proves little; trends come from running the same prompts on a schedule.

How to do it

1. Write 10 to 50 prompts in your buyers' words: category ("best payroll software for a 20-person agency"), problem, comparison and branded prompts, each tagged by type.
2. Choose the engines and markets your buyers use, and report each engine separately.
3. Run every prompt on a fixed schedule, with the same setup each time.
4. Log each run: the brands named in order, how the answer describes you, and the cited URLs.
5. Report rates over a fixed window, such as 30 days: AI visibility, share of voice, first mention and citation gaps.

Keep the prompt list stable within a period so trends compare the same questions. How to track brand mentions in AI search results walks through a manual setup with a sample sheet. PromptSignal runs each prompt every 48 hours by default in each engine you track.