Policy and guidelines on the use of generative AI models in posts on Virological.org
Version 1, October 7, 2026
Generative AI tools can be genuinely useful in scientific writing — helping to tidy up prose, debug code, or get a first draft of a methods description off the ground. We have no interest in discouraging that. What we do ask is that you treat AI tools the way you would treat any other tool in your workflow: with calibrated scepticism, and a clear understanding of where they can and cannot be trusted.
Principle: you are responsible for everything you post. AI tools might hallucinate references, misrepresent uncertainty, and can produce fluent, confident-sounding text that is misleading or simply wrong. This is particularly consequential in genomic epidemiology and infectious disease research, where a careless interpretation could result in ill-judged public health decisions. Before anything goes into a post — whether AI-assisted or not — you should be able to defend it on the basis of your own expertise (e.g., which method was used and why).
Reproducibility: AI outputs are not deterministic, meaning the same prompt will not reliably produce the same result twice. This means AI cannot itself be a step in a reproducible analytical pipeline although it could generate a script, workflow or pipeline that will allow replicable results. Any code generated with AI assistance should be reviewed, tested, and shared as you would any other code, so that others can inspect and replicate the work independently.
Authorship: The text of the post should be the authors’ own work. An AI model cannot be listed as an author.
Transparency: For posts where generative AI was used please include a statement at the top stating that the post follows these guidelines. Explicitly state what AI was used and how.