第一，养成在写作过程中记录AI使用的习惯。你在哪个环节用了AI、用了什么工具、做了什么修改，记录下来。投稿声明的时候直接拿这些记录来填，不用回忆。
第二，定期关注目标期刊的AI使用政策更新。出版社的政策在快速变化，你半年前看到的政策可能已经更新了。投稿前去期刊官网确认一下最新要求。
不同出版商对AI使用的政策是不一样的。比如：
Elsevier allows authors to use generative AI and AI-assisted technologies to refine and edit manuscripts prior to submission, provided this use occurs with appropriate oversight and disclosure. Authors who use these tools must include a required declaration statement at the end of the manuscript immediately above the references or bibliography; however, basic checks of grammar, spelling, and punctuation do not require a declaration. These tools must not be used as a substitute for human critical thinking, and authors remain responsible and accountable for the work, including reviewing and editing the content as needed after using the tool.
Karger permits authors to use generative AI/LLM tools when preparing a manuscript, including for language refinement. If such a tool has been used as part of a study or manuscript, the use must be clearly declared in the manuscript (in the Methods section, or in the Acknowledgements section when the article type does not include a Methods section) and cited in line with Karger’s software citation policy. Authors must guarantee the accuracy and originality of the manuscript and must include details on how the accuracy of any generative AI-based output was verified.
Nature permits authors to use an LLM or other AI tool for AI-assisted copy editing of human-generated text to improve readability and style and to correct errors (including wording and formatting changes). This type of AI use does not need to be declared, provided it remains limited to copy-editing and does not involve generative editorial work or autonomous content creation. In all cases, human accountability for the final version of the text is required, and authors must agree that any edits reflect their original work.
第三，投稿前跑一次AI检测。现在检测工具越来越准，你越早知道自己的论文AI率是什么水平，越有时间从容应对。
第四：搞清楚投稿踩坑与合规红线：
严禁 AI 署名：AI 不具备人类问责能力，绝对不能作为作者或共同作者列入论文。
严禁 AI 生成或篡改数据与图像：特别是实验结果图（如 WB 条带、电镜图、流式图等），使用生成式 AI 处理视觉证据会被视为严重学术不端并导致撤稿。
保护数据隐私：切勿将未发表的实验数据、机密信息或手稿上传至公共 AI 平台，以免数据被用于模型训练或泄露。
人工严格复核：AI 容易产生“幻觉”（如编造不存在的文献），所有 AI 生成的引文、数据和结论必须逐一人工核对。
投稿前，建议务必查阅目标期刊的 Guide for Authors 中关于 "Artificial Intelligence" 或 "AI Disclosure" 的最新具体政策，以免因格式不符被退稿。
AI检测的大方向很明确：工具越来越准、政策越来越细、审稿人越来越关注。与其被动应对，不如提前把功夫做在前面。把声明做实，把检测过了，把核心内容确保是自己的独立思考。
这四件事做到位了，不管政策怎么变，你都不慌。
NES耐思润色，为您匹配相同领域英语母语编辑进行深度人工润色，确保您的英文表达母语化，为您稿件的成功投稿保驾护航！
