How BrandScope measures AI visibility
Every week we run your brand through a preset set of questions across 6 AI assistants (ChatGPT, Perplexity, Gemini, Qwen, GLM, Kimi). We record whether — and how — you appear, then turn that into a single AI Health Score.
The AI Health Score (0–100)
A weighted composite of 5 factors:
| Factor | Weight | What it measures |
|---|---|---|
| Presence | 30% | AI coverage — how many of the 6 AIs mention you (at least once) |
| Accuracy | 25% | Context accuracy — how often AI describes you correctly vs your positioning |
| Wins | 20% | Competitiveness — % of recommendation questions where you rank #1 |
| Authority | 15% | Independent source rate — % of cited sources not on your own site |
| Freshness | 10% | Source age — median age of cited sources (coming soon) |
If a factor has no data for a run (e.g. Accuracy before judging completes), it uses a neutral value so the score isn't dragged down unfairly.
The 5 factors in detail
Presence — AI coverage
For each of the 6 AIs, we check whether your brand appears in any answer. Presence = % of AIs that know you. A low score means some AIs don't know you exist yet.
Accuracy — Context accuracy
When an AI mentions you, we judge whether its description matches your actual positioning (accurate / partial / misleading). This catches AI saying wrong things about you — the mistakes you never see.
Wins — Competitiveness
In "best tool for X" style questions, we record where you rank. Wins = % of those questions where you're #1. This shows whether AI recommends you when a decision is being made.
Authority — Independent source rate
When an AI cites sources, we check whether they're third-party (not your own site). AI trusts brands backed by independent coverage (Product Hunt, reviews, directories).
Freshness — Source age
The median age of cited sources. AI that quotes 5-year-old threads gives stale advice — this factor (coming soon) will flag it.
Sampling & why numbers can move
AI assistants don't give identical answers every time — there's natural variance (we've measured ±10% run-to-run). To reduce noise:
- Each (question × AI) is sampled multiple times and averaged (2–3× on Growth and above).
- Trend matters more than any single week — the report shows your history, not just one snapshot.
So: don't overreact to a single week's number. Watch the trend.
Data sources
- 6 AI assistants — ChatGPT (GPT-5.6), Perplexity, Gemini, Qwen, GLM, Kimi (via OpenRouter).
- 5 preset questions per week — discovery, comparison, and objection questions your customers actually ask.
- Sampling — each (question × AI) run multiple times, averaged for stability.
- Raw data — every AI answer is visible in the Mention details section, so the score is always auditable.
Questions? Email support@brandscope.dev.