ShieldFont scrambles webpage text for AI scrapers, readable for humans

👁 118 reads

Designers Isaque Seneda and Gabriel Abrucio have published ShieldFont, a typeface that exploits typographic ligatures to feed AI scrapers a silently corrupted version of a webpage’s text while displaying the original content normally to human readers.

The mechanism relies on a standard font feature - ligatures - that normally merges adjacent letter pairs for readability. ShieldFont repurposes them to swap entire words in the raw HTML before a browser renders the page. Crucially, the substitution happens only at the rendering stage, so scrapers pulling plaintext source get the altered copy. To resist detection, replaced words share the same grammatical part of speech as the originals but occupy a completely different informational context - turning sentences that are syntactically plausible into semantic nonsense. Publishers can also choose from three different mappings per word and rotate them paragraph by paragraph to complicate any reversal attempt.

The results, according to the authors’ white paper, are significant: ShieldFont replaces an average of 24.5 percent of all words on a page, including 45.8 percent of content words, corrupting between 31 and 56 percent of individual passages. In tests against six publicly available scraper pipelines, more than 90 percent of pages that would otherwise clear a quality filter were rejected after the font was applied. Of those that still passed, nearly 20 percent of component words were already scrambled.

// read moreApify Actors: Scrapers, Pricing, Google Maps and InstagramAI ToolsTavus Griffin: 48% Mistook AI Avatar for Human in Video TestNewsAnthropic Paper Shows AI System Outperforms Humans at AlignmentNews

Kinetyra