How The Answer Index is measured
Rankings are never for sale. Client firms are always disclosed. Corrections within 48 hours.
1. What is measured
The index measures which firms AI engines actually name when asked the questions real clients ask. Four engines: ChatGPT, Gemini, Perplexity and Google AI Overviews. Each edition covers one market and practice area with its own question set and firm universe.
2. The question set
Each edition tracks a fixed question set (typically 300) spanning practice areas, matter types and location phrasings, drawn from real search demand data for the market. Questions are weighted equally. The set is frozen once live: any change requires a version bump noted in the changelog, since changing questions breaks month-over-month comparability. Exact prompt strings are private: publishing them would let firms optimize for the test instead of the market and would contaminate the engines being measured. Journalists and researchers may review the full set under NDA by writing to team@aeoengine.ai.
3. Collection
Every question is run against every engine during the measurement window (days 1 to 3 of the month), producing the edition’s answer corpus. Runs use clean sessions with no account personalization. Answer text is stored unmodified for audit.
4. Entity resolution
Each firm has a maintained entity record: legal name, d/b/a brands, common misspellings and domains. Mentions are matched against this table. Firms may submit corrections to their entity record at any time.
5. Scoring
Share of voice: a firm’s total mentions divided by all tracked firm mentions in the corpus. Appearance rate: the share of questions naming the firm at least once. Ranking is by share of voice. DR and estimated organic traffic are pulled from Ahrefs at measurement time and displayed as context only. They never affect rank.
6. Eligibility
The firm universe is built per edition from firms actively marketing the practice area in the market, assembled from directories, advertising presence and market research, plus any firm that requests measurement. Minimum universe size 100. Being a client of AEO Engine neither includes nor excludes a firm.
7. Client disclosure
Some measured firms are AEO Engine clients. Every client firm is tagged on the board. Measurement and parsing run identically for every firm, client or not.
8. Limitations
Generative engines are probabilistic: results vary by phrasing, session and time. Each edition is a snapshot of the window, not a permanent standing. Month-to-month movement below 0.5 points should be read as noise. Single market, single practice area per edition. The interval between the inaugural edition and the second is approximately five weeks; every interval after that is an exact calendar month.
9. Corrections and changelog
Errors reported to team@aeoengine.ai are investigated and resolved within 48 hours. Every correction is logged here with date and description.
- 2026-07-28 (National 100 · All Practice Areas): Inaugural August 2026 edition published. National 100 question set frozen at v1.
- 2026-07-28 (Personal Injury 100 · National): Inaugural August 2026 edition published. Personal Injury 100 national universe.
10. No sponsorship
The index carries no advertising and no sponsors. Rankings are never for sale. The index is funded entirely by AEO Engine’s operating business.