Scholarlink

Artificial intelligence (AI) is increasingly used in medical writing due to its ability to automate content creation, improve structure, and save time. However, concerns remain about scientific accuracy, reference reliability, and the detectability of AI-generated content. This review aims to describe AI use in medical writing and the effectiveness and reliability of AI detection tools. A systematic review was conducted following PRISMA 2020 guidelines. PubMed, Web of Science, and Scopus were searched for English-language original articles published between January 2021 and June 2024. Inclusion criteria were original studies focusing on AI in medical writing. Data were extracted on AI tools, accuracy, plagiarism, writing efficiency, and AI detection performance. Of 585 articles, 27 met inclusion criteria. Surgical fields were most represented (52%). ChatGPT-3.5 (49%) and GPT-4 (27%) were the most used tools. AI-generated texts were fluent and structured but often contained factual errors and fabricated references. While AI-assisted writing improved efficiency, human revision remained essential. Detection tools like GPTZero, Originality.ai, and Sapling performed well for fully AI-generated texts but struggled with partially AI-assisted content. Human reviewers showed limited ability to identify AI outputs. AI enhances efficiency and structure in medical writing but faces challenges in accuracy and originality. Both AI detection tools and human reviewers are unreliable, highlighting the need for ethical standards, transparency, and sustained human oversight.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial Intelligence in Medical Writing: A Systematic Review of Current Practices

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nouhaila Charef,
  • Somaia El Ghazi,
  • Noura Qarmiche,
  • Hind Bourkhime,
  • Mohammed Omari,
  • Samira El Fakir,
  • Nada Otmani

摘要

Artificial intelligence (AI) is increasingly used in medical writing due to its ability to automate content creation, improve structure, and save time. However, concerns remain about scientific accuracy, reference reliability, and the detectability of AI-generated content. This review aims to describe AI use in medical writing and the effectiveness and reliability of AI detection tools. A systematic review was conducted following PRISMA 2020 guidelines. PubMed, Web of Science, and Scopus were searched for English-language original articles published between January 2021 and June 2024. Inclusion criteria were original studies focusing on AI in medical writing. Data were extracted on AI tools, accuracy, plagiarism, writing efficiency, and AI detection performance. Of 585 articles, 27 met inclusion criteria. Surgical fields were most represented (52%). ChatGPT-3.5 (49%) and GPT-4 (27%) were the most used tools. AI-generated texts were fluent and structured but often contained factual errors and fabricated references. While AI-assisted writing improved efficiency, human revision remained essential. Detection tools like GPTZero, Originality.ai, and Sapling performed well for fully AI-generated texts but struggled with partially AI-assisted content. Human reviewers showed limited ability to identify AI outputs. AI enhances efficiency and structure in medical writing but faces challenges in accuracy and originality. Both AI detection tools and human reviewers are unreliable, highlighting the need for ethical standards, transparency, and sustained human oversight.