01The tracking code really does not run
Until consent, the browser neither downloads nor executes it — the address is set aside rather than deleted, because consent has to be able to put it back.
02It also catches what other plugins add
Some plugins paste their tracker straight into the page, where the usual mechanism cannot see it. Those are matched by their address instead — found on a live store, where a Google Ads tag was running under a banner that believed it had blocked it.
03Consent is tied to the text that was shown
The policy version is a hash of the open categories, the banner text and the inventory. Change any of them and the old consent stops counting, so the question is asked again.
04Proof without identifying anybody
The server keeps counts of (day, policy version, choice) and an archive of every policy text. Person-level proof would mean giving the visitor a durable identifier — opening a new processing operation in order to prove consent for the old one.
05Reject is as easy as accept
Both buttons carry the same class and reject is printed first. A test asserts their class attributes are identical, because consent obtained through a harder refusal is not freely given.
06Nothing is cached wrong
No cookie, no nonce, no logged-in check on the front end. A consent banner that varies the page per visitor breaks full-page caching, and a nonce dies behind it silently.
07Legal pages from your own inventory
The cookie policy is generated from the categories and rows you actually filled in. An empty category prints no invented line — declaring a processing operation you do not run is worse than declaring too little.
08The banner never depends on the licence
If it went quiet when payment lapsed, the site would start writing tracking cookies without consent — and the fine lands on the site owner, not on us.
09A hundred and twenty-one services it can name
The library recognises 121 third-party services by the addresses they load from, together with the cookies each one sets and how long those last. Thirty-three of the records are complete; the other eighty-eight carry a visible “verify this” mark instead of a plausible guess, because a cookie table that invents a retention period is worse than one with a gap in it. Payment and bot-protection scripts are recognised too but never offered as blockable — holding one back breaks a checkout silently.
010A scan that runs itself and reports only what changed
Put the front-end audit on a weekly or daily timer and it tells you when a new unheld third-party script or an unlisted cookie appears — not the whole list again. Switching the schedule on is Pro; switching it off and scanning by hand are free.