Academic AI disclosure requirements depend on how the tool shapes your intellectual contribution, not just whether you used AI at all.
When Disclosure Is Required
You must disclose AI use when it directly generates core intellectual content — drafting arguments, synthesizing literature, interpreting data, or writing substantial portions of your manuscript. Most journals now require explicit statements if tools like ChatGPT or Claude produced text that appears in your paper, even if you edited it afterward. The key test: if removing the AI's contribution would fundamentally change your paper's reasoning or conclusions, disclosure is mandatory.
What Qualifies as Routine Editing
Standard editing assistance typically doesn't require disclosure. This includes grammar checkers, spell-checkers, translation tools for non-native speakers, and citation formatters. Many journals explicitly exempt these "writing assistance tools" from disclosure requirements. If you're using AI purely to polish existing text you wrote — fixing awkward phrasing, checking verb tense consistency, or formatting references — that generally falls under routine editing.
The Gray Zone
The boundary gets fuzzy with advanced AI features. Using AI to paraphrase your own draft raises questions: are you just improving clarity, or is the AI restructuring your argument? Similarly, AI-powered literature summarization tools exist on a spectrum. Reading a paper and asking AI to explain a confusing methods section is different from having AI synthesize multiple papers into your literature review. When in doubt, err toward disclosure.
Practical Guidelines
Check your target journal's author guidelines — policies vary significantly. Nature requires disclosure for any "substantive" AI use in writing. PLOS asks authors to describe AI's specific role. Some journals ban generative AI for text entirely but allow it for code or image analysis.
For citation verification specifically, tools that automatically check reference accuracy against published metadata fall clearly into routine editing territory, since they're validating your work rather than generating new content.
The emerging consensus: transparency protects you. A brief methods note explaining exactly how you used AI demonstrates rigor rather than suggesting impropriety.
