Real answers, never simulated
We query the 8 engines with the questions your customers ask, on a schedule, and extract mentions, citations and rankings from the actual responses.
Methodology
How we measure AI visibility — with attribution
No black boxes. Every score traces back to real answers.
Five-dimension scoring
Technical, content, authority, presence and sentiment — each dimension is derived from observed behaviour, not opinion.
Treatment vs. control
Lift is measured with paired prompt sets and reported with confidence intervals.
Failure codes
Each issue carries a failure code, so fixes are ranked by the lift they are expected to produce.
- 01
Anchored scoring
Scoring bands are computed by rules; the LLM may only fine-tune within a band — never free-scored.
- 02
Real engines only
Zero simulated data. Every metric traces back to an actual scan of a live engine.
- 03
Paired control queries
Control and treatment prompt pairs are baked into the data model via pair_id.
- 04
Three-tier attribution
Hard / strong / observed — language is graded by confidence, never overclaimed.
- 05
Engine-version aware
Engine version-change dates are tagged and excluded so upgrades never pollute attribution.
- 06
Auditable to the answer
Any score drills down to the original engine response — no score is a black box.
Run one free check to see exactly how AI talks about your brand.
Send us the question your brand should own — eight engines answer on the spot, with actionable optimization steps.
one prompt · eight engines · thirty minutes