V0.1 / METHODOLOGY
How we measure brand visibility in AI
We probe the major LLMs with niche-specific prompts and translate their answers into one number you can act on — the ECHO.
ECHO = 50 × 30 × 20
01 — 03The ECHO Index is a weighted blend of three things every brand needs from AI search:
Citation
How often the LLM actually names your brand in its answer. Zero mentions = zero ECHO no matter what else you do.
Sentiment
Whether the mention is positive, neutral, or negative. Being known and hated isn't a win.
Authority
Whether the LLM cites your own domain as a source. If yes — you control the narrative.
What we scan
01 — 06Every audit fans out across all active providers in parallel. Every prompt × every model = one row in the result matrix.
- 01
OpenAI
GPT-5
EN / RU
- 02
Anthropic
Claude Opus / Haiku
EN / RU
- 03
Google
Gemini 2.5
EN / RU
- 04
Perplexity
Sonar Pro
EN
- 05
GigaChat
Sber
RU
- 06
YandexGPT
Yandex Cloud
RU
Russian engines (GigaChat, YandexGPT) are first-class — most Western tools skip them.
“If AI doesn’t cite you, you don’t exist in the answer.”
What happens during an audit
01 — 05- 01
Intake
We fetch your homepage and parse <title> + meta description, so the LLMs are graded against your real positioning — not their hallucination of the domain name.
- 02
Multi-provider fan-out
We send the same prompts to every active provider in parallel — full matrix, no cherry-picking.
- 03
Extraction
We extract brand mentions, sentiment heuristics, and cited source URLs from each raw response.
- 04
ECHO computation
Citation, sentiment, and authority are aggregated and weighted 50/30/20 into one number from 0 to 100.
- 05
Snapshot
Each ECHO computation is persisted as a time-series snapshot so you can see trends after weekly re-scans (Phase 3).