摘要：
目的 评估AI政策制定相对自主的（简称：政策自主型）OA医学期刊生成式AI政策的覆盖水平与薄弱环节，为中国科技期刊分阶段建立和完善AI政策提供参考。方法 以DOAJ收录的609种政策自主型OA医学期刊为数据来源。基于ICMJE、COPE、EASE等国际指南构建涵盖6个维度18项指标的二元评分体系，由两名研究者独立评分，Cohen’s κ中位数为0.85。结果 364种（59.8%）期刊未制定任何AI相关政策；有政策的245种（40.2%）期刊平均得分为7.05分（满分18分）。政策内容集中于作者端文本披露，“限制AI用于图像”“AI用于数据规范”“审稿人/编辑AI限制”“违规处理条款”等指标覆盖率偏低。仅12种（2.0%）期刊达到高分标准（≥12分）。高分期刊的共同特征为基础披露要求完整、结构化声明较清晰、审稿人与编辑使用限制更明确，但在数据规范、违规处理和动态更新方面仍存在明显盲区。结论 政策自主型OA医学期刊的AI政策整体处于起步阶段。中国科技期刊可采取起步阶段-进阶阶段-完善阶段的分阶段建设路径，优先明确作者披露与责任归属，再逐步纳入图像、数据、同行评审和违规处理等高风险环节，最终实现流程嵌入、动态更新和能力提升。
关键词:
生成式人工智能,
学术出版,
开放获取,
医学期刊,
AI政策,
出版伦理
Abstract:
Purposes To assess the coverage level and weak points of generative AI policies among OA medical journals with relatively autonomous policy formulation， and to provide references for Chinese scientific journals to establish and improve AI policies in stages. Methods A total of 609 OA medical journals with relatively autonomous policy formulation indexed in DOAJ were used as the data source. Based on international guidelines from ICMJE， COPE， EASE， and other organizations， a binary scoring system comprising 6 dimensions and 18 indicators was developed. Two researchers independently performed the scoring， with a median Cohen’s κ of 0.85. Findings Of the 609 journals， 364 （59.8%） had no AI-related policy. Among the 245 journals （40.2%） with AI policies， the mean score was 7.05 out of 18. Policy content was mainly focused on author-side textual disclosure， while indicators such as “restrictions on AI use for images” “standards for AI use in data” “restrictions on AI use by reviewers/editors” and “violation-handling provisions” showed low coverage. Only 12 journals （2.0%） met the high-score threshold （score ≥12）. The common features of high-scoring journals included comprehensive basic disclosure requirements， clearer structured statements， and more explicit restrictions on AI use by reviewers and editors； however， substantial gaps remained in data-related standards， violation handling， and dynamic policy updating. Conclusions AI policies among OA medical journals with relatively autonomous policy formulation are generally still at an early stage. Chinese scientific journals may adopt a staged development path consisting of an initial stage， an advanced stage， and an improvement stage： first clarifying author disclosure requirements and responsibility attribution， then gradually incorporating high-risk areas such as images， data， peer review， and violation handling， and ultimately achieving workflow integration， dynamic updating， and capacity building.
Key words:
Generative artificial intelligence,
Scholarly publishing,
Open access,
Medical journal,
AI policy,
Publication ethics
田甜, 倪明, 王琳辉, 叶娟英, 王尔亮. 开放获取医学期刊生成式AI政策现状与启示——基于AI政策制定相对自主期刊的分析[J]. 中国科技期刊研究, 2026, 37(6): 864-873.
TIAN Tian, NI Ming, WANG Linhui, YE Juanying, WANG Erliang. Current status and implications of generative AI policies in open access medical journals： an analysis of journals with relatively autonomous AI policy formulation[J]. Chinese Journal of Scientific and Technical Periodicals, 2026, 37(6): 864-873.