"How visible is my business to AI?" deserves a better answer than a consultant's gut feel or a tool that scores you 68 on Monday and 74 on Wednesday with nothing changed. The whole value of a score is that its movement means your movement. That requires determinism: same site in, same number out. Every time.
AnvixaAI's AI Readiness score is exactly that: a 0–100 computed by code from stored, inspectable signals about your site. No model in the scoring loop, no randomness, no drift. Here's what's actually inside it.
The four signals
Schema coverage. Of the pages that matter on your site, how many are serving published structured data right now? A perfect document on one page doesn't make a machine-readable business; coverage is the foundation, and pages you've deliberately set aside as ignored don't count against you.
Entity completeness. Does your knowledge graph actually describe a business? People with roles, services with descriptions, locations with real addresses, a profile with contact facts. The score rewards a graph an assistant could answer questions from, not a stub with a name and a logo.
Content quality. This one is measured in an unusual place: not your prose, but the JSON-LD your pages actually produce. Do service pages yield real service nodes? Do articles carry article markup, FAQs their question structure? It's outcome-based: grading what machines will actually receive, not what your pages look like.
Install & serving signals. Is the snippet live, verified, and actually delivering schema to visitors? Published documents that never reach a real page load are worth zero, and the score says so.
The AI Report tab: one honest number on top, and the four signals that explain it underneath.
Why determinism matters more than it sounds
A deterministic score gives you three things a fuzzy one can't. Causality: if the number moved, something real changed: you published pages, completed an entity, or your site drifted. Accountability: every sub-score traces to inspectable facts, so "why is content quality 74?" has an actual answer. Progress you can trust: when the score goes from 61 to 76 after a week of work, that delta is yours, not measurement noise.
The score also refreshes itself. The platform re-crawls your site on a cycle, recomputes readiness from the fresh signals, and the number quietly tracks reality. That includes downward, honestly, if pages drift or something breaks. A score that can't go down isn't a score; it's a compliment.
What a number can't do alone
A score tells you where you stand, not what to do Tuesday morning. That's the second half of the AI Report tab: every weakness the score detects is turned into a concrete fix, priced in points by simulating the fix and re-scoring. Which is its own article, but the principle matters here: a score you can decompose is a score you can act on.
- Same site in, same score out: no model, no noise, no Wednesday surprises
- Four signals: schema coverage, entity completeness, content quality, install & serving
- Content quality grades the JSON-LD your pages actually produce: outcome, not effort
- Recomputed automatically after every re-crawl, downward included