Illustration of a decoy font shielding text from AI using optical illusions in a digital interface.

Innovative Decoy Font Shields Text from AI Scraping Through Optical Illusions

User avatar placeholder
Written by Sam Johnson

2026-07-31

In an era where AI scraping raises privacy concerns, Decoy Font offers a creative shield against unwanted data collection. Created by Mixfont, this typeface employs optical illusions to confound AI chatbots, preserving messages for human eyes only.

Decoy Font enabling humans to hide their writing from AI detection.
Decoy Font lets humans hide their writing from AI

Deciphering the Double Vision

Decoy Font leverages a clever optical trick: it combines a foreground of thin, clear outlines with a background of blurred dark shapes. Users can read both layers by altering their focus or distance from the screen, yet AI systems typically only pick up the crisply outlined foreground text. As Eric Lu, associated with Mixfont and an AI researcher, explains, this design interrupts AI’s ability to accurately interpret the text.

Mixfont's decoy font tricking AI tool ChatGPT through visual deception.
Mixfont used the font to fool ChatGPT

From Ghost Font to Decoy Innovation

Building on earlier experiments like “Ghost Font,” which relies on animated illusions, Decoy Font takes a more static approach. This allows it to sidestep limitations faced by its predecessor, such as needing movement to be effective against AI. Ghost Font’s reliance on motion is highlighted by its experimental design, unlike the versatile and practical Decoy Font intended for traditional viewing.

Gibberish decoy message appearing suspicious but effective against AI scraping.
A gibberish decoy message also works but prompts suspicion

Proven Trickery Against AI Models

Lu’s experiments with Decoy Font showed impressive results. When tested on advanced AI models like GPT-5.6 Sol and Gemini 3.5, the AI misinterpreted images that humans clearly read as “sleep early” as “binge shows.” This highlights the font’s capability to mislead machines while humans see the intended message.

User encoding messages on the Mixfont website platform.
Users can encode their own messages on the Mixfont site

The Technique Behind the Illusion

The Decoy Font employs a “hybrid image” tactic, merging low and high spatial frequencies to play tricks on AI perception. This technique is reminiscent of famous illusions like an image of Marilyn Monroe that looks like Albert Einstein up close. The font uses this same principle to keep AI systems at bay.

For another fascinating take on blending creativity with practicality, explore how Anoma’s Revolutionary A1 Prehistoric Watch merges ancient techniques with modern design.

Credit image from Playtype and S&A related to decoy font design.
(Image credit: Playtype / S&A)

Shielding Messages with Style

Beyond its fascinating implementation, Decoy Font taps into broader discussions about AI’s reach and privacy. While initially a way to explore AI limitations, the project now also serves as a benchmark to monitor AI’s rapid advancements. As Lu notes, today’s unreadable font might soon be deciphered by advancing AI, making tracking its evolution crucial.

Comparing to ShieldFont’s Approach

Meanwhile, the Brazilian creative studio Seneda & Abrucio and the type foundry Playtype have introduced “ShieldFont,” another tool in the typographic arsenal against AI scraping. Unlike Decoy Font’s optical illusions, ShieldFont disrupts AI interpretation by altering HTML through OpenType glyph substitution, presenting altered sentences to AI without affecting human readability.

ShieldFont modifies website text source code, making words invisible to AI.
ShieldFont changes select words in the source code behind the text shown on a website (Image credit: Playtype / S&A)

Open Source for All

ShieldFont functions as an open-source tool, encouraging widespread adoption and adaptation in diverse typefaces. It offers a practical anti-scraping solution while propelling discourse on digital ethics and AI data harvesting. Its creators stress the importance of maintaining control over one’s own creative content in an age of data exploitation.

Text altered by ShieldFont remains readable to humans, bypassing AI filters.
The altered text remains coherent so it’s more likely to pass AI quality filtering and enter training datasets (Image credit: Playtype / S&A)

Tools like Decoy and ShieldFont underscore the ongoing dialog between creativity and technology, advocating for the protection of human work in the digital realm.

Both designs represent a fusion of creativity and defense, leveraging typography’s traditional role as a vessel for ideas while adapting to ward off technological encroachment.

Dive into the world of unique innovations with the Innovative Clock Using Water, which challenges traditional time-telling methods.

Sources & Links

Source: dezeen.com

Frequently asked questions

How does Decoy Font protect against AI scraping?

Decoy Font uses optical illusions by combining a foreground of thin, clear outlines with a background of blurred shapes, confounding AI chatbots and preserving messages for human eyes.

What is the technique behind Decoy Font’s illusion?

The Decoy Font employs a ‘hybrid image’ tactic, merging low and high spatial frequencies to play tricks on AI perception, similar to famous illusions like an image of Marilyn Monroe looking like Albert Einstein up close.

How is ShieldFont different from Decoy Font?

Unlike Decoy Font’s optical illusions, ShieldFont disrupts AI interpretation by altering HTML through OpenType glyph substitution, presenting altered sentences to AI without affecting human readability.