A pair of designers is proposing a font-based defense against AI scraping that keeps web pages readable for people while making the underlying text less useful to automated collectors.
Ars Technica reports that ShieldFont relies on ligatures, a standard font feature that can replace character sequences with different visual forms. In normal typography, ligatures improve how letters look together. Here, the idea is to display readable words to humans while leaving altered or nonsensical text in the HTML that a scraper might collect.
The project is intended as a practical opt-out tool for publishers who do not want their work used in unauthorized AI training. It joins a growing set of technical and legal responses to large-scale web scraping.
The limitation is that this is not a universal block. Sophisticated crawlers may adapt, and the technique depends on how scrapers extract text. Still, it shows how resistance to AI training is moving beyond robots.txt and lawsuits into the design of web content itself.