Ismael Said Criado retweetledi
I developed an ethical AI use minicourse that attracted more than 1,500 people.
I’ve read a lot of weak takes on AI in academic writing.
This paper is different.
10 high-level principles every researcher should take from this paper on ethical AI use in academic writing
It does not say “use AI for everything.”
It does not say “avoid AI completely.”
It draws a clearer line:
AI can support academic writing.
It should not replace scholarship.
——-
Here are the 10 biggest lessons
1. AI assistance is not the same as AI authorship
The paper is clear that generative AI tools do not meet accepted standards for authorship.
A tool cannot take responsibility for a paper.
Human authors still carry that responsibility.
2. Transparency is non-negotiable
If AI is used in the writing process, its use should be clearly and explicitly disclosed.
3. Use AI where it is strongest
The authors place grammar, spelling, readability, and translation in the most acceptable tier of use.
4. Treat idea-generation uses with caution
The paper places outlining, summarizing, improving clarity, and brainstorming in a middle category.
These uses can be reasonable, but only if the author already has the core ideas and carefully reviews the output.
5. Do not outsource your thinking
The paper warns against using AI to generate new concepts, draft de novo text, interpret data, conduct literature reviews, check ethical compliance, or judge plagiarism.
Those are the parts of writing in which scholarship itself is tested.
6. Accuracy still belongs to the human author
LLMs generate fluent text from statistical patterns.
They lack built-in fact-checking, which is one reason hallucinations and bias remain real problems.
7. References are a major danger zone
The paper cites evidence that AI-generated references can be fabricated or inaccurate at alarming rates.
That is why authors remain fully responsible for checking every citation and source.
8. Ethical use should preserve human skill, not erode it
One of the strongest points in the paper is that scholars should worry not just about productivity, but about what happens when people become too dependent on AI for ideation, writing, and early interpretation.
That kind of dependence can weaken scholarly growth.
9. Human oversight is not optional
The paper leans on three core ideas for responsible use: human vetting, substantial human contribution, and acknowledgement.
In practice, that means AI output should always be reviewed, checked, and owned by the researcher.
10. Before using AI, ask four hard questions
The paper ends with a practical checklist.
Before using AI in a manuscript, ask:
→ Are the primary ideas and interpretations still mine?
→ Am I using AI in a way that preserves core research and writing skills?
→ Have I checked everything for accuracy, reliability, and bias?
→ Have I disclosed clearly how AI was used?
That, to me, is the strongest takeaway from the paper.
Reference:
Cheng A, Calhoun A, Reedy G. Artificial intelligence-assisted academic writing: recommendations for ethical use. Advances in Simulation. 2025;10:22. doi.org/10.1186/s41077…
English
