International Panel Sets Clear Rules for AI in Medical Writing
A multidisciplinary team from hospitals and universities across China, Italy, Japan, Canada, and Australia has published the first unified framework for using artificial intelligence in medical research. The guidelines, released in March 2026 in the journal Regenesis Repair Rehabilitation, draw a sharp distinction between permitted and prohibited AI tasks-ending the confusion that has left researchers unsure whether using AI crosses an ethical line.
The problem was clear. Major publishers had adopted AI policies, but they contradicted each other. One journal demanded every prompt and software version; another simply said "be honest." Meanwhile, large language models routinely fabricate citations, inventing authors, journal names, and DOIs with confidence. Feeding unpublished patient data into public chatbots created privacy risks. Researchers needed practical guidance, not conflicting rules.
What the Framework Prohibits
AI cannot generate or manipulate primary research images. Microscopy scans, gels, and flow cytometry plots must remain untouched by generative tools. References are off-limits-the fabrication risk is too high.
AI also cannot invent missing steps in the Methods section or alter actual data in the Results section. The core science must come from genuine experiments.
What AI Can Do
Generative AI gets approval for grammar correction, abstract refinement, and discussion summaries-as long as a human verifies every claim. Workflow diagrams and non-data visuals are permitted with full disclosure. The authors themselves used ChatGPT and Gemini during writing, but only for language and structure, never for scientific content.
The framework includes section-specific guidance. In Methods, AI can improve readability but cannot add missing steps. In Results, text can be polished, but numbers must come from actual experiments.
Transparency Over Secrecy
The authors rejected a policing approach. "We are not the AI police," they said. "The worst outcome would be researchers secretly using AI anyway, afraid to admit it."
The real danger is invisibility. Using AI to rewrite a discussion or suggest citations is acceptable-provided it appears in a disclosure table. Claiming AI-generated text as original thought or a fabricated reference as real crosses into misconduct.
"Transparency turns a black box into a simple tool," the team said.
Practical Benefits
Researchers now have a checklist to share with collaborators or include in supplementary files. Journals gain a ready-made audit trail to streamline peer review. Non-native English speakers can use AI for language support without ethical worry.
The framework removes guesswork and shifts the conversation from "is AI cheating?" to "how do we use AI well?" The authors committed to updating the rules as AI evolves, protecting science from silent errors while embracing writing tools that improve efficiency.
For more on AI applications in research, explore AI Research Courses.
Your membership also unlocks:
