Science (AAAS) AI Policy: From Total Ban to policy-required disclosure

Science's current AI policy has five components:

1. AI can't be an author. Like every major journal, Science doesn't allow AI tools to be listed as authors, and sources cited in Science journal content may not be authored or coauthored by AI tools. The reasoning: authors must take accountability for published work, and AI can't do that.

2. Some AI use must be declared, and where depends on the use. The Science journals' AI usage guidelines set the rule use by use:

• Writing or editing code in a way that changes functionality, and data generation or analysis, must be described in the methods section, with model name and version number.
• Writing or drafting manuscript content, translation of manuscript text, and editing or formatting of visualizations of research data or results must be declared in the acknowledgments section, with model name and version number.
• Refinement, correction, editing or formatting to improve clarity of language, literature searches, formatting of data reported in the manuscript, refinement or formatting of code that leaves its functionality unchanged, and assisting with gathering references need no declaration.

The editorial policies add that authors who use AI-assisted technologies as components of their research or as aids in the writing or presentation of the manuscript should note this in the cover letter and in the methods section or acknowledgments section.

3. AI-generated images require editor permission. AI-generated images and other multimedia are not permitted in the Science journals without explicit permission from the editors. Exceptions may be granted in certain situations, such as images or videos in manuscripts specifically about AI or machine learning. They are evaluated case by case and should be disclosed at the time of submission.

4. Presenting AI-generated content as original research is not allowed. The guidelines list presenting AI-generated content as though it were original research from non-AI sources as not allowed.

5. Authors stay accountable. Authors are accountable for the accuracy of the work and for ensuring that there is no plagiarism. They must also ensure that all sources are appropriately cited and should carefully review the work to guard against bias that may be introduced by AI. Editors may decline to move forward with manuscripts if AI is used inappropriately.

Science's current guidelines are specific about which uses are allowed and which declarations are required, but they are less specific about penalties:

• The editorial policies say editors may decline to move forward with manuscripts if AI is used inappropriately.

• The 2023 editorial said that a violation of the policy at that time would constitute scientific misconduct no different from altered images or plagiarism of existing works. The current guidelines do not restate that penalty.

• The Science journals handle integrity concerns, corrections, expressions of concern and retractions following COPE guidelines.

This doesn't mean Science will retract your paper if you forgot to mention you used Grammarly. The policy targets substantive undisclosed use, like drafting entire sections of text without the required declaration. The sensible reading is that you should treat the declaration rules as binding.

Science's declaration rules mean you need to think about AI use before you even begin the submission process. The examples below are illustrative wording, not text from Science.

In the cover letter:

"During preparation of this manuscript, we used Claude (Anthropic) to assist with drafting the Results and Discussion sections. All AI-assisted text was reviewed, verified, and edited by the authors, who take full responsibility for the content."

In the acknowledgments (for writing, drafting or translation):

"The authors acknowledge the use of Claude (Anthropic, Claude 3.5 Sonnet) to draft portions of the manuscript text, which the authors then revised."

In the methods (for AI used in the research):

"Analysis code for the image segmentation was written with the assistance of Claude (Anthropic, Claude 3.5 Sonnet). The authors reviewed all code and verified its output against the original data."

The key principle: be specific about which tool, which version, and which parts of the work were affected.

Science's image policy is worth understanding because it distinguishes between kinds of figures:

• Not allowed: Generating, refining, correcting, editing or formatting images, diagrams or other figures for illustrative purposes only. Exceptions must be discussed with the editor.

• Allowed with a declaration: Editing or formatting of visualizations of research data or results, declared in the acknowledgments with the model name and version number.

• Possible with editor permission: AI-generated images in manuscripts specifically about AI or machine learning, evaluated case by case and disclosed at submission.

The editorial policies also say Science does not allow selective manipulation of any part of a digital image. For a standard biology, chemistry, or physics paper, don't expect the editor to approve AI-generated figures.

Science's editorial policies say it uses the Crossref Similarity Check service to detect plagiarism and that editors may use Proofig to screen images. The policies do not describe a dedicated AI-detection step or an AI checkbox in the submission system, so the declaration depends on authors being accurate.

Post-publication, the Science journals handle integrity concerns following COPE guidelines, which can lead to corrections, expressions of concern and retractions. Readers with concerns can contact science_data@aaas.org.

Science's rules depend on the use of the tool, which means you need to match each use in your workflow to the right declaration. Whether this is appropriate vigilance or excessive bureaucracy depends on your perspective, but either way, you need to comply.

Before writing:

• Decide upfront whether you'll use AI tools. If yes, plan your disclosure language early.

• Keep records of which tools you used and which manuscript sections they touched. You'll need this specificity for the declarations.

During writing:

• Use AI tools if they genuinely help, but don't use them just because they're available. Drafting, translation and code that changes functionality create a disclosure obligation.

• Don't have AI generate your cover letter, the document where you're supposed to note AI use. The irony won't be lost on editors.

During submission:

• Note AI use in the cover letter and place the declaration where the guidelines say for each use.

• Be honest about the extent of AI involvement. "Used for minor language editing" when AI generated entire paragraphs is the kind of discrepancy that can become an integrity case.

Common pitfalls:

• Forgetting the cover letter note. Most authors remember the manuscript sections but overlook that Science's editorial policies also say to note AI use in the cover letter.

• Using generic disclosure language. "AI tools were used during preparation" isn't specific enough. Name the tool and its version, and describe the use.

• Assuming code generation never counts. If AI wrote or edited code in a way that changes functionality, describe it in the methods with the model name and version.

Not sure if your manuscript meets Science's disclosure requirements? A Science submission readiness check can help you identify gaps before submission.

Science's AI policy applies to all journals in the Science family:

• Science

• Science Advances

• Science Translational Medicine

• Science Signaling

• Science Immunology

• Science Robotics

The editorial policies are written for all Science journals, and the AI usage guidelines are titled "Science journals' guidelines for AI usage". If you're submitting to Science Advances thinking the rules might be softer, they aren't.

Comply proactively if:

• You used any AI tool (ChatGPT, Grammarly, Copilot) during manuscript preparation

• The journal requires AI use declarations in the methods or acknowledgments

• Your institution has its own AI use policy that may be stricter

Less concerned if:

• You used AI only for language refinement (Science's guidelines say no declaration is needed)

• Your use was limited to literature searches or gathering references (no declaration is needed)