Every researcher using AI tools is navigating a policy landscape that’s still being written. What was acceptable at your institution six months ago may have changed. What’s acceptable in one department may differ from another. Understanding the current state prevents career-limiting mistakes.
The Policy Landscape in 2025
German research universities are generally in a disclosure-and-acknowledge model: AI use is permitted, disclosure is required. DFG issued guidance in 2023 stating that AI tools cannot be listed as co-authors, that authors are responsible for AI-generated content in their submissions, and that undisclosed AI use in grant applications constitutes a violation of good scientific practice. Similar frameworks govern ERC and Horizon Europe funding.
What Disclosure Actually Looks Like
Check your target journal’s AI policy before submission — most major publishers (Elsevier, Springer, Wiley, Nature, Science) have explicit policies updated as of 2024. Standard disclosure language looks like: “The authors used [tool name] to [specific purpose, e.g., improve language clarity in the methods section]. All scientific content was developed and verified by the authors.” Be specific — “we used AI” is insufficient; state which tool and for what purpose.
The Authorship Question
AI cannot be listed as an author. Authorship requires the ability to take responsibility for the work — AI cannot sign off on corrections, respond to reviewer critiques, or stand behind claims about data integrity. This is consensus among major publishers and funding bodies. If your institution or co-authors ask about this, the answer is clear: no AI authorship, with human responsibility for AI-assisted content.
Coursework vs Research
Policies for coursework (seminar papers, assignments) are often stricter than policies for research publications and grant writing. Check your university’s Prüfungsordnung for AI-specific provisions — many German universities added explicit rules in 2023–2024. A violation in coursework can result in Täuschungsversuch (deception attempt) findings with serious academic consequences, separate from research misconduct rules.
The Safe Default
When in doubt about whether a specific AI use is permitted: use AI for structural and mechanical assistance (summarizing, improving clarity, code suggestions), disclose all AI use explicitly, write all substantive intellectual content yourself. This approach satisfies current guidelines at virtually all major research institutions and is defensible if questioned.
