Learn AI Today
Subscribe

Home›Watermarks & Detection

Watermarks & Detection

How AI watermarking works, and why AI detectors are not the same thing.

Watermarks and Detection covers the two ways people try to tell whether a text was written by AI, and why they should not be confused. A watermark is a hidden pattern that a model adds as it writes, which only the company holding the key can test for. An AI detector guesses from writing style, and it can be wrong about human writing.

Read the full introduction

The explainer covers how Claude's watermark works, when it started and what editing does to it. The comparison of watermarks and detectors is written for anyone who has been accused of using AI, or who has to judge such a claim, and sets out what each kind of evidence can and cannot show. The third post explains why translating your own writing with Claude can leave a mark on words you wrote yourself.

Detection is changing quickly, so each post carries the date it was last checked.

Recent posts