Because AI answers vary between sessions, engines and phrasings, no single answer is a reliable measurement. A prompt portfolio fixes the questions, the engines and the schedule so results can be compared across periods, the same way a survey fixes its questionnaire.
What a good portfolio covers
- Informational intents: what is the category, how does it work.
- Comparison and alternatives: X vs Y, alternatives to Z.
- Best-provider intents: best agency, tool or service for a specific need.
- Problem-solution intents: how do I fix or achieve something the brand solves.
- Local and language variants for each market that matters.
- Reputation intents: is X trustworthy, reviews of X.
Each run records which brands are mentioned and recommended, which domains are cited and what claims are made. Aggregated, those records become mention rate, recommendation rate, citation rate and competitor share of voice. The portfolio itself should stay stable between periods, with new prompts added in versioned batches so trends remain comparable.