Skip to content

News

AI detection, and what a flag is evidence of.

Updated 2 stories

HumanizersFor teachers

An AI Humanizer Rewrites Text to Dodge Detectors and Tends to Damage the Prose

An AI humanizer rewrites machine-written text so a detector will not flag it. The mildest version swaps a single word; the most aggressive rewrites whole passages. Pangram Labs names three techniques on its own blog: synonym swaps, inserted nonsense phrases, and plain damage to spelling and grammar [1]. A 2025 audit of 19 such tools found that humanizers tend to degrade the writing, to varying degrees, and the paper's tier averages put the original ahead in every tier [2].

Bill Nguyen & Human

Detectors

0 stories

No stories yet.

Pangram

0 stories

No stories yet.

Humanizers

2 stories

No stories yet.

Classroom

0 stories

No stories yet.

Policy

0 stories

No stories yet.

Writing

0 stories

No stories yet.

Writers

Bill Nguyen

AI detection in education: what a detector reads, what a flag is evidence of, and what a school does with one.

Detectors · Pangram · Humanizers · Classroom · Policy · Writing

The detection beat: how AI-text detectors work and where they fail, what the published studies measured, what Pangram, Turnitin and the rest say about themselves, and what a teacher, a student or a school does when a flag lands. The question underneath every piece is what a reading is evidence of and what it is not. It offers no way to prove who wrote a document, and no number that stands for a person.

Start with a paper you know.

Paste something you know a person wrote, such as your own writing, and see what our detector marks in it. It is free to start, with 2,000 words a day, or 1,000 papers a month with a verified school email.