Every number, explained.
AuditOps sells one thing: that the findings are real. So the model is published, including where it stops. Nothing is a trade secret, and the code that implements it is the same code described here.
How a score is built
Every category starts at 100 and pays for each finding it owns. A high-severity finding costs 20 points, medium 10, low 4, and informational nothing, because an informational finding is a check that passed. Findings are deduped by category and title before scoring, so a problem present on ten pages is charged once and the page count lives in the finding title.
The overall score is weighted
The overall is a weighted mean of the seven categories, not a flat average. SEO and performance carry 1.25, AI visibility and security 1.00, accessibility and responsiveness 0.90, and link health 0.70. The weights sum to 7.0, so the overall stays on the same scale as a simple average and older reports stay comparable. Grades: A+ from 95, A from 90, B from 80, C from 70, D from 60, F below.
Measured or flagged
A finding is labelled measured when the crawler observed the value directly, and flagged when it is a judgement a human should confirm. Every finding carries the evidence it came from, so you can check the work rather than trust it.
What this is not
It is not a penetration test, not a WCAG conformance audit, and not a Lighthouse run. It reads what your server sends to a crawler and reports what is provably there or provably missing. If your site renders its content in the browser rather than on the server, most of it is invisible to this crawler and to the LLM crawlers that matter for AI visibility.