Here’s the rewritten article:
**The Battle Against AI-Written Content: Wikipedia’s New Plug-In to Make AIs Write More Like Humans**
Ever tried to spot a piece of text written by an AI? It’s not as easy as you’d think. But now, thanks to a new open-source plug-in, the game’s afoot. Siqi Chen, a tech entrepreneur, has created a plug-in for Anthropic’s Claude Code AI assistant that tells the AI to dial down the robot speak and write more like a human.
The idea behind this plug-in, called Humanizer, came from the Wikipedia community’s AI Cleanup project. French Wikipedia editor Ilyas Lebleu founded the project, which has been working since late 2023 to identify patterns of AI-generated content. In August 2025, they put out a comprehensive guide to the language and formatting patterns that give AI writing away. Chen’s Humanizer plug-in is based on this guide, and it’s designed to make Claude’s output more “human-like”.
But what does it actually do? In simple terms, Humanizer is a “skill file” that provides a list of instructions for Claude to follow. Instead of a standard prompt, the skill file is formatted in Markdown, a language that Claude’s been trained to understand. The result? The AI’s output sounds less robotic and more relaxed.
But how effective is it? We tested it out, and while it’s definitely a step in the right direction, it’s not perfect. Some of the directions in the plug-in can lead to unexpected outcomes, depending on the task at hand. For instance, one of the directions tells the AI to have opinions and react to them – not exactly what you want if you’re writing technical documentation.
The Wikipedia guide itself is full of examples of AI writing, from over-the-top language to awkward phrases. Humanizer aims to teach Claude to replace these with plain facts, along with example transformations to show how it can do it better.
But the tricky thing about AI writing detection is that it’s not foolproof. As we’ve written before, there’s nothing inherently unique about human writing that makes it distinguishable from AI-generated content. One reason for this is that AIs can be prompted to avoid certain language patterns, just like the Humanizer plug-in does.
**The goal of Humanizer is to help AIs write more like humans, and it’s an interesting step forward in this area. But the battle against AI-written content is far from over.**
