New Font Aims to Thwart AI Data Scrapers by Altering Hidden Text

Isaque Seneda and Gabriel Abrucio have released a novel typeface called ShieldFont, designed to combat large-scale web scraping for AI training. The font uses ligatures to replace entire words in the HTML source, but the browser renders the intended readable text for visitors. This means that automated tools that download the page’s raw code encounter altered, meaningless content, effectively blocking their data collection.

The designers describe the tool as a practical way for site owners to opt out of unauthorized AI training. They argue that merely swapping words with synonyms or antonyms would be too easy for sophisticated scrapers to undo, while replacing text with random gibberish could be flagged and bypassed, so ShieldFont aims for a balanced solution. The white paper accompanying the launch positions the font as a defensive measure against the growing practice of AI companies harvesting public web data.