This is a replay. Every request, verdict and page was recorded from the real prototype server. It leaves out the live server and the Shield Font, and the page is shown finished, not drawn by the font.
Disclosure is an extension of the concept of a Shield Font (shieldfont.org). It is a test of whether a web page can give correct text only to requests that carry a verifiable signed declaration, without changing what people see. In the test, signed requests read 100% of the words and unsigned requests read 61%. Screenshots are correct in both cases. No real agents were tested, and OCR and public font mapping still expose text.
The value exchange of web scraping has shifted from driving traffic via search to pulling content and information away from site-holders.
Site-holders, such as publishers and creators, want to control who reads their content, and users need their tools to keep working. Existing controls are either easy to get around or block everyone, including search engines, screen readers and people who copy and paste.
It matters because AI agents and scrapers read the web at scale, and blocking them shouldn't cost real users access.
Instead of blocking all automated readers or none, Disclosure gives correct text only to requests that carry a verifiable signed declaration. The page looks the same to people, and a screenshot is correct either way. Unsigned requests get degraded text from the server.
Building on the Shield Font idea, this gives site-holders a finer control than an on/off switch, while search, screen readers and copy-paste can keep working for requesters who declare who they are. The broad approach is to encourage declaration rather than simply punish bad behavior.
If a page can give correct text only to requests that declare who they are, without changing what people see, then sites get finer control and requesters have a reason to declare. If that holds, declaring who you are becomes a normal part of how software reads the web.