It doesn't read your code. It shops.
An agent opens your store in a real browser, adds a product, reaches checkout, and judges what it saw against 115 rules. About three minutes. Nothing installed.
Four pages, one agent, one score
Opens your homepage
Checks it is a real store, finds a product the way a shopper would.
2reads: scripts off, then a real browserReads the product page
Price, shipping, reviews, variants, schema. Adds to cart.
40rulesOpens the cart
Totals, shipping estimate, express wallets, edit and remove.
25rulesReaches checkout
Walks to the payment step. Card never entered. Nothing ordered.
50rules103 rules for humans, 12 for AI agents
Every rule is pass, partial, fail, not applicable or not assessable. Rules it could not judge, and rules your store has nothing for, are dropped and never guessed. Each failing rule gets a revenue-at-risk band and a fix.
Find the leak. Fix it. Prove it.
A finding is worth nothing until the score moves. Re-audit the same store and we show which rules flipped from failing to passing.
Watch plans ›Could not judge, so not scored
If bot protection stops the walk at checkout, its rules are dropped from the score. Never guessed.
Coverage next to every score
You always see how many of the rules actually ran on your store.
Ranges, not point estimates
Revenue at risk is a band, estimated as a percentage of revenue, with the source named.