The three parts
The score is a weighted sum. Checks are half of it: the things a crawler can verify on your pages, like titles, canonicals, headings and structured data. Speed is thirty percent, read from real page-load measurements. Presence is the last twenty percent, and it covers how findable the brand is off its own site.
- Checks, weighted 0.5: what a crawler can verify on the page.
- Speed, weighted 0.3: measured load performance.
- Presence, weighted 0.2: how the brand appears beyond its own site.
What we ask the models
Four assistants get asked about the brand: GPT, Claude, Grok and Gemini. For each one we record whether it recognised the company at all, how it described it, whether that description was accurate, and which competitors it named instead. A model that confidently describes the wrong company is a different problem from one that has never heard of you, and the report separates them.
Where the score is incomplete
Presence has two halves and only one of them currently has a source. The AI half is measured. The search half is not: Google closed whole-web programmable search to new engines in January 2026, so there is no source we can read for where a brand ranks, and a paid search API is a decision we have not made. Presence therefore reads low for everyone, including us. We would rather say that than quietly report a half-measurement as a whole one.
Real scored examples
Every published report is a real scan rather than a sample. These three were run the same way yours would be.