I’ve written quite a few things using AI. After finishing, I’d read it over and thought it flowed pretty well. But when I sent it out, the other person replied, “This was written by AI, right?” My heart skipped a beat. I had clearly changed a few words and adjusted the tone, but they still spotted it right away.
This isn’t your fault. It’s the “default mode” of AI writing that’s to blame. When large language models generate text, they choose the phrasing that “holds true in the most situations.” They write for everyone, not just for you. These “default choices” leave a trail of clues. Wikipedia editors have compiled these clues into a guide. The seven tools I’m introducing today are specifically designed to tackle these issues, and they’re all Agent Skills that can help you remove the “AI-like” feel from your English writing.
A Roundup of 7 Tools for Removing AI-Generated Content
User Experience
I've been using these tools for quite a while, and I've picked up a few tips along the way:
When using tools, start by running a check with the detection tool—don’t rush to rewrite the entire piece right away. It will highlight where the text sounds “AI-generated,” so I’ll know exactly where the problems are before making changes. Tackle one layer at a time: first, remove clichés; then adjust sentence length; and finally, refine the tone. Trying to do too much at once can easily make the text sound disjointed.
These tools are used in combination. “stop-slop” removes clichés and adverbs, “humanizer” adjusts the rhythm and adds details, and “avoid-ai-writing” performs the final review. The approach varies depending on the text style: I use `academic-humanizer` for academic texts, `avoid-ai-writing` for blog posts and marketing copy, `taste-skill` for front-end copy and design specifications, and `De-AI-Prompt-Enhancer` for Chinese prompts. These tools don’t replace me—they’re my assistants.
I always keep the facts intact. I don’t let the tool generate numbers, nor do I let it change people’s names. If the tool is missing information, I fill in the gaps. I also control the extent of the rewriting: if I change too much, the original meaning gets lost; if I change too little, it still sounds like it was written by an AI. I read the text aloud, shortening parts where I stumble and leaving parts that flow smoothly as they are.
Make your prompts as specific as possible. Tell it who the target audience is, what tone you want, and which words to avoid—this will significantly improve the output. Finally, read through it yourself; the tool can’t take responsibility for me. My goal isn’t to fool a detection tool—I want the text to sound like a real person is speaking. By varying sentence structures and pacing, it reads as if I’m speaking myself, and the “AI-like” quality naturally fades away.