Scholarlink

Aim <p>Patient education is important in managing rheumatological diseases. One potential resource is the use of artificial intelligence (AI)-generated educational materials; however, further research is needed. This study assessed how ChatGPT, Google Gemini, Microsoft Copilot, Apple Intelligence, and Meta-AI created patient guides for systemic lupus erythematosus, rheumatoid arthritis, osteoporosis, polymyalgia rheumatica, and fibromyalgia.</p> Subject and methods <p>The AI-produced guides were evaluated for readability through the Flesch-Kincaid grade level and ease score, their reliability was tested using the modified DISCERN scale, and originality was verified using the QuillBot plagiarism tool. Two experts evaluated the outputs using validated tools. The statistical analysis involved comparisons at a significance level of <i>p</i> &lt; 0.05.</p> Results <p>Significant differences were observed among the AI platforms. Google Gemini produced the most readable guides, with the lowest grade level and highest ease score, but also had the lowest reliability and the highest similarity (plagiarism) percentage. Microsoft Copilot achieved the highest reliability score but had the lowest readability. All AI-generated materials exceeded the recommended sixth-grade reading level. Plagiarism rates varied, with Meta AI showing the lowest and Google Gemini the highest similarity percentages.</p> Conclusion <p>The materials created by AI tools present compromises between their readability, reliability, and originality, as no platform achieves the highest scores in all these areas. Human supervision remains crucial for AI-based patient education systems to maintain accuracy and ethical standards while matching patient literacy levels. Future research should evaluate a wider range of medical conditions and advancements in AI models.</p>

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

Evaluating readability, reliability, and originality of artificial intelligence-generated patient education guides for common rheumatological conditions

  • 期刊论文

  • 发表日期：

    2025年11月22日

查看全文
  • Jyothis G. Saji,
  • Jaziya Jabeen,
  • Joseph T. Antony

摘要

Aim

Patient education is important in managing rheumatological diseases. One potential resource is the use of artificial intelligence (AI)-generated educational materials; however, further research is needed. This study assessed how ChatGPT, Google Gemini, Microsoft Copilot, Apple Intelligence, and Meta-AI created patient guides for systemic lupus erythematosus, rheumatoid arthritis, osteoporosis, polymyalgia rheumatica, and fibromyalgia.

Subject and methods

The AI-produced guides were evaluated for readability through the Flesch-Kincaid grade level and ease score, their reliability was tested using the modified DISCERN scale, and originality was verified using the QuillBot plagiarism tool. Two experts evaluated the outputs using validated tools. The statistical analysis involved comparisons at a significance level of p < 0.05.

Results

Significant differences were observed among the AI platforms. Google Gemini produced the most readable guides, with the lowest grade level and highest ease score, but also had the lowest reliability and the highest similarity (plagiarism) percentage. Microsoft Copilot achieved the highest reliability score but had the lowest readability. All AI-generated materials exceeded the recommended sixth-grade reading level. Plagiarism rates varied, with Meta AI showing the lowest and Google Gemini the highest similarity percentages.

Conclusion

The materials created by AI tools present compromises between their readability, reliability, and originality, as no platform achieves the highest scores in all these areas. Human supervision remains crucial for AI-based patient education systems to maintain accuracy and ethical standards while matching patient literacy levels. Future research should evaluate a wider range of medical conditions and advancements in AI models.