ACM Policy on Authorship

The ACM Policy on Authorship governs every work ACM publishes — journals, transactions, magazines, and the proceedings of its 70-plus annual conferences and special interest group (SIG) events. It sets two things: who counts as an author, and, since ACM’s 2023 update addressing generative AI, what authors must disclose about AI-assisted content. Individual SIGs and conferences (SIGCHI, SIGCSE, and others) point authors back to this same publisher-level policy rather than maintaining separate authorship rules, though some add venue-specific guidance on top of it for generative-AI use in peer review itself.

Who ACM Considers an Author

Consistent with the norms most scholarly publishers apply (see CASRAI’s ICMJE entry for the most widely cited version of this test), ACM’s policy ties authorship to substantive intellectual contribution and accountability: an author must have helped originate or create the ideas or content of the work, made a substantial contribution to it, and remain accountable for the work and how it is presented. This is a contribution-and-accountability standard, not a labor-and-effort standard — running experiments, formatting a manuscript, or acquiring funding, on its own, does not meet it. For a broader comparison of how different fields define authorship contribution thresholds, see CASRAI’s guide, Types of Authorship in Research.

Generative AI Cannot Be Listed as an Author

ACM’s policy is explicit that generative AI tools and technologies, such as ChatGPT, may not be listed as authors of an ACM-published work. This follows the same reasoning nearly every major publisher and standards body has converged on since generative AI tools became widely available: an AI system cannot hold accountability for a work’s accuracy or integrity, cannot give consent to be listed, and cannot be held responsible for correcting or retracting it — all things authorship requires. See CASRAI’s AI as author entry for how this reasoning is applied across publishers generally, and generative AI for the underlying technology definition.

When Generative AI Use Must Be Disclosed

Where a generative AI tool is used to help create content — text, tables, graphs, code, data, or citations — ACM’s policy requires that use to be fully disclosed in the work itself. ACM’s own guidance is deliberately risk-averse on edge cases: if an author is uncertain whether a particular tool’s use needs disclosing, ACM’s instruction is to err on the side of disclosure rather than omit it.

The policy carves out one narrow exemption from this disclosure requirement: basic word-processing aids, such as standard spelling and grammar checkers, are not treated as generative AI use requiring disclosure. The line ACM draws is between tools that generate or substantially rewrite content (which must be disclosed) and tools that merely check or lightly correct an author’s own already-written text (which do not need to be). This mirrors the distinction CASRAI’s AI tool disclosure entry describes more generally across publishers: routine editing aids are typically exempted, while drafting, summarizing, code-generation, or image-generation tools are not.

How to Write the Disclosure Statement

ACM’s guidance places the disclosure in the work’s Acknowledgements section, and offers example wording authors can adapt, such as: “ChatGPT was utilized to generate sections of this Work, including text, tables, graphs, code, data, citations, etc.” In practice this means naming the specific tool used and describing, at the level of what was generated (a section of text, a code snippet, a figure), rather than a vague blanket statement that AI was “used somewhere.” CASRAI’s generative-AI disclosure statement entry has more on structuring this kind of statement so it holds up under editorial or post-publication scrutiny.

How This Compares to Other Publisher Policies

ACM’s ban on AI-as-author and its disclosure-with-a-narrow-exemption structure is broadly consistent with the approach taken by ICMJE-aligned medical journals and most major STM publishers: no publisher of note currently permits an AI system to be listed as an author, and disclosure of substantive AI-generated content is now close to universal policy, even where the specific required wording or location (acknowledgements vs. methods vs. a dedicated AI-use statement) differs. See CASRAI’s guides on NIH Authorship Guidelines and BMJ Authorship Guidelines for two other funder/publisher variants of the same underlying principle. Unlike the CRediT taxonomy’s structured, role-by-role contribution statement, ACM’s policy does not require a formal per-author contribution breakdown — disclosure is about the AI tool’s role, not each human co-author’s.

Examples

• A conference paper’s Acknowledgements section states that a named large language model was used to draft an initial literature-review paragraph, which the authors then revised and verified — this is a correctly disclosed use under ACM’s policy.
• A paper lists an AI writing assistant as a co-author in the byline — this violates ACM’s policy regardless of how the tool was used, because AI tools cannot be listed as authors at all.

Counter-Example

An author runs a finished, human-written manuscript through a standard spelling-and-grammar checker before submission and does not mention this in the Acknowledgements. Under ACM’s exemption for basic word-processing aids, this is not a policy violation — it falls outside what the disclosure requirement covers.