The AI Authority Index

What the AI Authority Index measures

AAI weights four separately measured pillars into one comparable score. Each pillar is scored from stored AI answers, so every point traces back to observable evidence.

• Discovery (30%) — How often AI systems surface your brand when buyers ask category and solution questions — before any brand name is mentioned.
• Recommendation (25%) — Being named is not being recommended. This measures how strongly AI puts your brand forward as an answer rather than listing it in passing.
• Coverage (25%) — Breadth across question types, buyer intents, models and markets. A brand that only appears for one narrow question is fragile.
• Positioning (20%) — How AI characterises your brand relative to competitors — including where its description of you is wrong or outdated.

Why counting mentions is not enough

• A mention is not a recommendation — assistants also name brands in passing and as counter-examples
• One answer is not a measurement — assistants are non-deterministic
• Breadth matters: appearing for one narrow question is fragile
• Accuracy is part of visibility: how AI describes you affects whether you are chosen

AI Authority Index bands

• Invisible (0–19) — AI systems rarely surface the brand — recommendation share is effectively absent.
• Emerging (20–49) — AI systems recognize the brand intermittently, but it is not yet a default recommendation.
• Established (50–69) — Recognized by AI systems and present in consideration sets, though not yet a leading recommendation.
• Strong (70–84) — Frequently recommended by AI systems across high-intent prompts; competitive but not dominant.
• Leading (85–100) — AI systems consistently treat the brand as a primary recommendation in its category.

Why asking ChatGPT once is not the same measurement

• Repeated measurement: every question is executed multiple times per cycle
• Multiple AI systems: ChatGPT and Gemini are measured separately so divergence is visible
• Entity normalisation: brand-name variants resolve to one identity
• Competitor benchmarking against the brands AI names alongside you
• Historical comparison: cycles are stored so movement is like-for-like
• Approximately 34 unique buyer-intent questions in the current canonical pack, each executed on ChatGPT ×3, Gemini ×2 — up to about 170 planned AI evaluations per cycle.

What to do with your AAI score

The full pipeline is documented in our methodology, and every term is defined in the Academy glossary.

• Read the pillar mix: low Discovery is a visibility problem, weak Recommendation is a positioning problem
• Prioritise the pillar with the largest weighted gap — Discovery carries 30% of the index
• Correct what AI gets wrong about your capabilities with clearer public evidence
• Track direction across cycles rather than day-to-day noise