‘Humanizer’ tool can erase signs of AI-written text — alarming scientists

2026-07-07 · Source: Machine learning : nature.com subject feeds · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Research Methodology & Innovation · Depth: Intermediate, quick

Summary

A new academic "humanizer" tool, released on June 20 by Jie Ding at the University of Minnesota, aims to personalize the tone of AI-written research papers and grant proposals by removing characteristic signs of AI usage. While some researchers, like Francisco Maria Calisto, praise its effectiveness for tasks such as emails and code documentation, others, including Miguel Angel Blazquez Rodriguez and Cassidy Sugimoto, express alarm, fearing it will encourage undisclosed AI use, which they consider deceptive and harmful to science. Max Spero, CEO of AI-detection platform Pangram, noted his system caught most "humanized" text but is developing upgrades. Ding maintains the tool is an editing aid, clarifying that non-disclosure is the ethical issue, not the tool itself, and has updated its GitHub site to reflect this, changing its description from "removes the usual AI tells" to "sharpens clarity and voice" and adding an ethics note. The tool works by instructing an AI to remove specific stylistic "tics" and add evidence for claims.

Key takeaway

For research scientists considering AI assistance in writing, you must prioritize transparency. While tools like the "humanizer" can refine AI-generated text for tone and clarity, your obligation to disclose AI use remains paramount. Failing to declare AI assistance constitutes misconduct, regardless of editing. Be aware that AI detection systems are also advancing to identify "humanized" content. Ensure your institutional and publisher guidelines are strictly followed to maintain academic integrity.

Key insights

A new "humanizer" tool aims to mask AI-generated text, sparking debate over its utility versus potential for academic misconduct.

Principles

• AI-generated text often has a discernible, repetitive style.
• Ethical issues stem from non-disclosure, not editing tools.
• AI detection tools evolve to counter "humanizer" techniques.

Method

The tool adapts a humanizer to academic content by instructing an AI to remove its "tics" (e.g., "not just X, but Y", em dashes) and add evidence for unsupported claims.

In practice

• Use "humanizers" to personalize AI-generated drafts.
• Apply specific stylistic rules to refine AI output.
• Be aware of evolving AI detection capabilities.

Topics

• AI Writing
• Academic Ethics
• AI Detection
• Text Humanization
• Research Integrity
• Natural Language Generation

Code references

• AIScientists-Dev/academic-humanizer

Best for: AI Scientist, Research Scientist, AI Ethicist

Related on AIssential

• 33 Tiny Habits That Make Your Writing Sound Like a Robot — AI-assisted writing often develops subtle, repetitive "tells" that betray its machine origin, which can be identified and corrected.
• On the troubling rise of generative AI suspicion in academic publishing — Overextended suspicion of GenAI in academic publishing leads to false positives and an "AI witch hunt."
• How to Humanize Your AI Writing in 10 Steps — Humanizing AI writing requires addressing both AI's common errors and incorporating distinct human writing techniques.
• Our Rules for AI Writing — Ethical AI writing involves using AI as an assistant, not a ghostwriter, to preserve human authorship, learning, and accountability.
• ‘Obvious markers of AI’: doubts raised over winner of short story prize — The definitive detection of AI-generated creative work remains elusive, posing challenges for literary awards and publishers.
• We’ve Reached the Point Where AI Has to Catch AI (The Internet Is Full of Slop) — AI detection tools are crucial for verifying content authenticity amidst widespread AI-generated text.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.