Policy for disclosure and credit allocation in LLM-assisted research

Develop and justify policy frameworks specifying how the use of large language models should be disclosed in research manuscripts and how authorship credit should be allocated for LLM-assisted works.

Background

The paper surveys heterogeneous journal policies ranging from bans on LLM authorship to encouragement of AI use with disclosure, and argues there are pragmatic reasons to consider how to credit LLM contributions. However, it does not endorse a specific policy.

The authors explicitly state that policy questions about disclosure and credit allocation remain, underscoring the need for clear, actionable standards to guide researchers and editors.

References

While the present paper has focused on the authorial status of LLM users in research contexts, there remain philosophical questions about the authorial status of LLMs and the ethics of using them, as well as policy questions about how LLM use should be disclosed and how credit should be allocated for works thereby produced.

— Authorship Without Writing: Large Language Models and the Senior Author Analogy  (2509.05390 - Hurshman et al., 5 Sep 2025) in Section V (Conclusion)

The analyses at the individual, collective, and industrial levels jointly indicate that generative AI has pushed the problem in creative domains from "can AI create?" to a more fundamental normative question: when creative agency is distributed among multiple human and machine actors, "to whom creative authorship belongs" becomes itself an open question (Uddin et al., 2025).

— What Makes Creation Human? Authorship, Reasons, and Meaningful Human Control in Generative AI  (2609.10738 - Cao, 9 Sep 2026) in Section 3, opening paragraph

Early in the discussions that lead to this paper, we brought up the impact of AI and how its use could be represented in contribution statements, but we did not come to conclusive results. One participant accurately described the current milieu by provocatively wondering, ``#1''{AI is increasingly used to contribute to many of these activities named in CRediT. Does that change anything? In one view, an author is making that contribution using a tool. In another view, the AI-enabled author is not doing the same epistemic work as before and is also using a system that is epistemically indebted to all of the content on the internet.}”

— Considering Contribution Statements in Visualization and HCI Research  (2608.12792 - Solen et al., 13 Aug 2026) in Section “Discussion,” subsection “Role of AI”