🌅 New Decoy Fonts Fool AI

SOURCE

Seneda & Abrucio (2026)

WHAT TO KNOW

• Brazilian creative studio Seneda & Arbucio and Danish type foundry PlayType have teamed up to create a font that prevents automated webpage crawlers from taking content without permission to train AI models. Called ShieldFont, the new protocol works by swapping out roughly half of all “content words”—nouns, verbs, adjectives, and adverbs—in a webpage’s source code with random, nonsensical words. When AI crawlers scrape the page’s text, the content appears incoherent and low-quality, resulting in more than 90% of the content being rejected by AI models’ quality filters. Clever code ensures the text displayed to humans is left alone.

WHY IT MATTERS

• ShieldFont represents the latest attempt by various designers and developers to create a font that fools AI and ensures content can’t be used to train large language models without permission. Digital font foundry Mixfont has released two “anti-AI” fonts in recent years. The first, called Ghost Font, uses thousands of tiny dots moving in opposite directions to make text appear as faint letters. The second, called Decoy Font, features foregrounded text with clear outlines layered over blurred background text that humans can still read.

CONNECT THE DOTS

• Both Decoy Font and Ghost Font rely on optical illusions to outsmart AI crawlers. The illusion used for Decoy Font is a modern version of hybrid imaging, which went viral in 2007 when MIT researchers released a portrait that showed both Marilyn Monroe and Albert Einstein, depending on the distance of the viewer.