Anti-AI fonts break accessibility and won't stop scrapers anyway

Fonts that scramble or obfuscate text to defeat AI scraping fail on two fronts at once. First, they wreck accessibility: screen readers and assistive tools parse the same mangled characters, so the technique locks out disabled readers—exactly the humans it claims to serve. Making an obfuscated page accessible would require machine-readable metadata, which only reappears as a gatekeeping problem, pushing toward centralized identity verification and, in effect, registries of disabled users. That is a privacy and civil-liberties dead end, not a fix.

The defenses are also technically doomed. Every published font, tech demo, and discussion thread hands AI vendors a fresh benchmark to train multimodal models against, and most existing schemes are already broken. The underlying logic is simple: if a person can read the text, a machine can eventually parse it. Elaborate motion-graphic or video-based tricks are too impractical for everyday web use to matter, and any obfuscation that did catch on would just draw more effort to crack it.

The author frames this as a cat-and-mouse game with a bad equilibrium. A web where content is routinely obfuscated demands expensive systems to read it—legitimately or not—and the natural endpoint is copy-protection, paywalls, and easier censorship, all cutting against the open web’s founding premise of free access to information. There is no silver bullet, AI capabilities aren’t going to regress, and the realistic baseline is that anything publicly viewable will eventually be machine-readable. Plan around that rather than fighting it with broken fonts.