Physical Review Letters AI Policy: ChatGPT and Generative AI Disclosure Rules for Physical Review Letters Authors

APS asks for disclosure in the figure caption (figure generation), the methods (AI used to conduct the research) or the Acknowledgment statement (other uses). Acceptable disclosure language for Physical Review Letters submissions, adapted to the right location:

"For our broad-impact physics advance communicable across physics subfields within 4 pages-focused manuscript at Physical Review Letters, we used ChatGPT-4o (OpenAI, version dated October 2024) to polish English-language phrasing in the Introduction and Discussion sections. We did not use generative AI for data analysis, figure generation, or substantive manuscript content. All authors reviewed and edited the AI-assisted text and take responsibility for the final manuscript."

Or, for AI-assisted code:

"For this Physical Review Letters submission addressing broad-impact physics advance communicable across physics subfields within 4 pages, initial Python code for the Bayesian regression analysis was drafted with Claude 3.5 Sonnet (Anthropic, version dated December 2024). All code was reviewed, modified, and validated by the authors before use; the final version is available at [repository URL]. Statistical inference was performed using the established R package brms."

What does NOT pass Physical Review Letters's desk-screen:

• For Physical Review Letters addressing broad-impact physics advance communicable across physics subfields within 4 pages: "AI tools were used in manuscript preparation." Too vague against the APS disclosure standard, which asks for the AI tool name and version, how the AI assisted, and how the authors directed and verified the output

• "We acknowledge AI assistance." (Too vague: APS asks for the tool name and version, how the AI assisted, and how the authors directed and verified the output.)

• "ChatGPT helped write this paper." (Insufficient detail on use case)

• No disclosure when AI was used substantively (APS says any substantive use must be disclosed in the paper itself)

For Physical Review Letters-targeted manuscripts, the patterns below are common AI-policy risk areas to check against the publisher's current guidance before submission.

AI disclosure missing despite obvious AI-assisted phrasing. Substantive AI-assisted drafting without a required disclosure can trigger an editorial query. Check whether your manuscript reads as AI-assisted

AI disclosure placed in the wrong manuscript location. APS puts figure-generation disclosure in the figure caption, disclosure of AI used to conduct the research in the methods, and other disclosures in the Acknowledgment statement. A misplaced disclosure can create an avoidable submission query. Check whether your AI disclosure is in the right section

Generic disclosure language without tool name and version. APS asks authors to state the AI tool name and version, how the AI assisted, and how the authors directed and verified the output. "AI tools were used" without specifics does not meet that standard. Check whether your AI disclosure has the required specificity

Source: Manusights internal review of Physical Review Letters-targeted submissions, 2025 cohort.

APS's AI policy framework continues to evolve as 2026 brings new ICMJE recommendations, COPE guidance refinements, and journal-specific clarifications. Physical Review Letters authors targeting broad-impact physics advance communicable across physics subfields within 4 pages submissions should track three signals throughout 2026:

Policy updates from APS. The publisher's public AI policy guidance is updated over time; APS says it will continue to monitor AI's capabilities and limitations and update the guidelines as the technology evolves. Physical Review Letters authors who pre-register their disclosure language at submission time tend to face fewer revisions during the 2026 transition period than authors who write boilerplate disclosures.

Field-specific clarifications for broad-impact physics advance communicable across physics subfields within 4 pages. Different research domains see different AI use patterns. APS's FAQ lists what counts as substantive AI use (for example scientific reasoning, figure generation, code generation that materially affects results, and data analysis) and says light polishing needs no disclosure, for broad-impact physics advance communicable across physics subfields within 4 pages work as for any other. Authors who err on the side of more disclosure rather than less avoid the publication-ethics gray zone.

Reviewer disclosure norms. The APS policy already covers peer reviewers: reviewers must disclose in a note to the editor any AI use to help draft or edit a review beyond light editing, and may not upload manuscripts or reports to unrestricted AI tools.

• Manusights internal preview corpus (150+ Physical Review Letters-targeted manuscripts, 2025 cohort)

Competitor pricing and feature claims on this page reflect publicly listed information as of 2026-05-14. Pricing and features may change; verify against each vendor's current product page before decision-making.