*LLMs in scientific publishing: perhaps the bigger question is not whether we use them, but how we use them.*
A recent Perspective in *Frontiers in Genetics* makes an interesting argument: scientific authorship should be judged by *intellectual leadership*, not simply by who produced the sentences.
The researchers who define the problem, formulate the questions, design the study, interpret the evidence and take responsibility for the conclusions remain the authors.
An LLM can help with writing. It should not take over the thinking.
The paper makes four practical recommendations for LLM-assisted authorship:
*1. Transparency* Substantial use of LLMs should be disclosed.
*2. Responsibility* Human authors remain accountable for every statement, result and conclusion.
*3. Verification* References, factual claims, figures and numerical results must be checked before publication.
*4. Purpose* LLMs should improve scientific communication, not compensate for weak reasoning or simply increase publication output.
The authors go one step further.
They propose that journals integrate *standardised LLM-based verification into the submission process before peer review*. Automated checks could examine references, factual consistency, figures, data integrity and technical issues.
That would not replace peer review.
Instead, it could allow reviewers to spend more of their limited time on questions that require human scientific judgement: originality, significance, conceptual coherence and interpretation.
The underlying message is worth considering:
*Science advances because of better questions, better evidence and better explanations — not simply better-written sentences.*
The challenge, then, may not be to keep LLMs outside scientific publishing, but to establish clear boundaries around where human intellectual responsibility must remain.
Article: Kirov B, Marinchev A, Peykov S, Galabov V. (2026). *From prohibition to integration: LLMs in scientific publishing*. Frontiers in Genetics, 17:1947911.
doi.org/10.3389/fgene.2026.1947911