摘要：

目的 系统调查国内科技期刊作者在论文写作中使用AI工具的现状，为期刊编辑部和高校及研究机构制定科学合理的AI治理策略提供实证依据。方法 采用问卷调查法，以国内工程技术和自然科学领域科技期刊作者为研究对象，深入调查了其在论文写作中使用AI工具的现状、行为模式、伦理认知与政策期待等。结果 AI工具在论文写作中普及率近8成，集中于语言润色、翻译等辅助环节。“AI腔”与“幻觉”问题是主要短板。作者对基础辅助与学术不端行为已形成较明确的伦理分层共识，但对边界模糊地带的规范需求强烈。期刊政策知晓度严重不足，作者对声明机制、边界划定和分级治理抱有高度期待。结论 期刊编辑部应明确边界与红线、推行结构化声明与分级处理机制、加强政策宣贯、规范审稿AI使用、提供合规工具；高校与研究机构应出台内部规范、提供工具访问、加强伦理教育与技能培训、营造学术诚信文化；作者自身应坚守学术自主、主动声明、提升AI素养并参与共同体建设。三方应形成合力，共同构建AI时代学术诚信新生态。

关键词: 科技期刊, 人工智能, 论文写作, 学术诚信

Abstract:

Purposes To conduct a systematic investigation into the current situation of AI tools used by authors of domestic scientific and technological journals in their paper writing， in order to provide empirical evidence for the journal editorial departments and universities and research institutions to formulate scientific and reasonable AI governance strategies. Methods The questionnaire survey method was adopted. The authors of scientific and technological journals in the fields of engineering technology and natural sciences in China were selected as the research subjects. The current situation， behavioral patterns， ethical cognition and policy expectations of their use of AI tools in paper writing were deeply investigated. Findings AI tools are widely used in scientific research writing， with a penetration rate of nearly 80%， mainly focusing on auxiliary tasks such as language polishing and translation. The issues of “AI style” and “hallucinations” are the main weaknesses. Authors have formed a relatively clear ethical classification consensus on basic assistance and academic misconduct， but they have a strong demand for the regulation of ambiguous areas. The awareness of journal policies is seriously lacking， and authors have high expectations for the declaration mechanism， boundary delineation， and hierarchical governance. Conclusions Journal editorial office should clarify boundaries and red lines， implement structured declarations， strengthen policy dissemination， regulate the use of AI in review， and provide compliant tools； universities and research institutions should introduce internal norms， provide access to tools， strengthen ethical training， and create a prudent and open culture； authors themselves should adhere to academic autonomy， actively declare， enhance AI literacy， and participate in community building. All three parties should form a joint force to jointly build a new ecological environment for academic integrity in the AI era.

Key words: Scientific journals, Artificial intelligence, Paper writing, Academic integrity