One number, published in the open. Every check, every weight, every penalty — because a legibility score nobody can audit is a marketing claim wearing a number.
The version is stamped on every scan. A score produced under one version is never compared with a score produced under another.
Catalogue checks contribute 80%, schema markup 20%. Blocked AI crawlers are then subtracted — 6 points each, capped at 30. Nothing is rounded, curved or tuned to make a demo look better.
(80% × catalogue) + (20% × schema) − (6 × blocked crawlers, max 30)
Weights sum to 100 before the blend. Measured per product, averaged across the catalogue.
Under ~180 characters there is nothing for an agent to extract facts from.
Material, dimensions, composition. The fields agents filter and compare on.
Every variant needs a stable identifier or an agent cannot reference what it buys.
Blank type means your product joins no comparison set at all.
Three or more per product. How use-case and occasion queries find you.
No weight means no delivery estimate - you drop out of speed-filtered results.
Multimodal agents read images. No image is a hard skip on visual surfaces.
Default Title tells an agent nothing about size, colour or configuration.
Brand attribution. Feeds brand queries and Merchant Center matching.
Structured data on the product page, worth 20% of the composite.
Every AI agent your robots.txt refuses costs 6 points of the composite, to a maximum of 30. These are the ten agents checked.
80 and above. Agents can read, compare and cite the catalogue.
50 to 79. Some products surface; the rest are guesswork.
Under 50. Agents cannot reliably answer questions about the catalogue.