How the AI Index is built

For every category, each assistant is asked the question a real buyer asks — “What are the best {category}?” — phrased naturally, with no brand names supplied. Each assistant is asked multiple times per refresh, because AI answers vary run to run; a single sample is an anecdote, not a measurement.

All 8leading assistants, every refresh — no engine is skipped because it is inconvenient to reach:

ChatGPTClaudeGeminiPerplexityDeepSeekGrokCopilotGoogle AI

The exact model version behind every captured answer is recorded with it. When an assistant ships a new model, the Index shows the shift rather than hiding it.

Brands are ranked by their AI Recommendation Score: how widely a brand is recommended (how many of the 8assistants name it) and how often (its share of voice and mention rate across all captured answers), plus how often AI cites the brand’s own site as a source. Presence in real answers is the only input — there is no editorial weighting.

• A brand must be recommended by at least two different assistantsto be ranked at all — one engine’s quirk is not a market position.
• Every ranking keeps its receipts: the verbatim answers the assistants gave are stored and shown, so any position can be checked against the answers that produced it.
• Each monthly refresh is captured as an immutable snapshot. Movement arrows compare against the previous snapshot; past editions are never edited.

Some categories — banks, insurers, agencies, professional services — get genuinely different answers in different countries, so those categories are also asked as a local buyer would ask, for the United States, the United Kingdom, Australia and Canada (“What are the best {category} in {country}?”), using the local term where it differs. Global-brand categories like electronics and software answer the same everywhere and are not collected regionally.

A regional edition is published only when its ranking meaningfully differsfrom the global one — a different #1, or fewer than three-quarters of the top 10 in common. When a market simply agrees with the global answer, the global page stays the single canonical ranking. Regional editions refresh monthly, offset two weeks from the global refresh.

The Index is refreshed monthly across all assistants and categories. Every page shows its refresh date. Between refreshes the ranking does not move — a stable, comparable measurement beats a jittery live one.

The Index reports what AI assistants recommend — it is not an endorsement by CiteHawk, and it is not a review site. If an assistant is wrong about a category, the Index will faithfully show you that it is wrong. That is the point.