Methodology.
What we measure, why these four signals, and how they roll up into one number. The framework is open; every reading shows you the answers it was scored from.
How a reading is taken.
For every reading we probe each model the same way, then read four signals off how it responds. The probes themselves, and how we tune them, are part of our calibration and aren't published. What is always open: the four signals below, the weights we apply, the composite math, and, on every reading, the actual answers each model gave, so you can see the evidence your score is built from.
Four dimensions, scored end-to-end.
Each model's reading is the weighted sum of four field measurements. The weights sum to 100%.
Presence
Did the subject appear at all, across the probes?
Highest when the subject appears across all three answers. The direct answer carries the most weight; the two peer-set answers make up the rest.
Prominence
When the model mentioned the subject, how early in the response did the name appear?
Higher when the name lands early in an answer rather than buried near the end, averaged across the answers that mention it. Zero if the subject was never mentioned.
Accuracy
Did the model give concrete specifics or hedge with uncertainty?
Read from the direct answer by a calibrated LLM judge: it scores how specific and confidently stated the answer is — dates, titles, named organizations versus vague or hedged language. It judges the answer's qualities, not its truth in the world.
Sentiment
Was the framing around the mention positive, neutral, or negative?
A calibrated LLM judge reads how the whole direct answer frames the subject — positive, neutral, or negative. 50 is neutral.
Every model. Different weight.
Per-model scores roll up into the composite using these weights. We weight by real-world usage share, heavier on the models that actually answer your buyers' questions today.
ChatGPT & Perplexity together carry 42%, they're where most discovery queries land today. We renormalize the composite over whichever models are live at scan time, so an unavailable provider never flatters or punishes your reading.
The composite, computed.
For each model:
model_score = presence × 0.30
+ prominence × 0.25
+ accuracy × 0.25
+ sentiment × 0.20Then the composite reading:
composite = Σ (model_score_i × model_weight_i)
/ Σ model_weight_i
(renormalized over models live today)What we don’t claim.
- NoteAccuracy is scored by a calibrated LLM judge, not a fact-check. It reads how concretely and confidently a model describes you — which correlates with the model actually knowing you — and the judge is validated against a hand-labeled rubric set before any methodology change ships.
- NoteSentiment is scored by the same judge, reading how the whole answer frames you. It cleanly separates positive from neutral from critical framing; sarcasm can still slip past it.
- NotePer-model scores are stable within a week but not bit-identical. Treat the index as a calibrated reading, not a deterministic score.
- NoteAll seven providers (ChatGPT, Claude, Gemini, Perplexity, DeepSeek, Grok, Meta) are live and queried on every scan. If a provider is temporarily unavailable, its weight is renormalized out of the composite so a single outage doesn't skew your reading.
- NotePer-model scores cluster for very well-known subjects. If LeBron James, Sam Altman, or Beyoncé are scored, all seven models will land within a narrow range; that's the calibrated signal, not a bug. The spread widens for less-famous subjects where some models recognize the name and others hedge.
Run the procedure on yourself.
One free reading. Each model's verbatim answer about you is yours to inspect.
Run a free scan ↗