What you get
- A dated, shareable report with every check shown, its evidence attached, and no score you have to take on trust.
- Starter files you can copy:
llms.txt, organisation and product JSON-LD, and a robots.txt that adds only what is missing. - A list of what to fix first, ordered by what actually stops an agent.
What this is not
This report states what is mechanically checkable about your site: which machine-readable surfaces exist, what they say, and whether they agree with each other. It does not predict whether an agent will find, trust or recommend you, and it does not measure your traffic or your ranking. Those are different claims with different evidence. Anyone selling you a score without showing you the checks is selling you an opinion.
How the checks were chosen
Every consequential choice behind this tool is written down as it was made — the alternatives, the reason, what it cost, and what went wrong — and published at the decision corpus. It is generated from the same repository as the rubric, so it cannot be tidied up after the fact. If you want to know why a check exists or why we refuse to score you, the answer is in there.
The rubric is public
Every check is written out at the rubric, with the reasoning and whether it is grounded in published practice or is our own judgment. The rubric is generated from the code that runs, so the two cannot drift apart. The whole tool is open source.