Comprehensiveness, Accuracy, and Readability of Exercise Recommendations Provided by an AI-Based Chatbot: Mixed...

Area of Science:

• Exercise Science
• Artificial Intelligence in Healthcare
• Digital Health

Background:

• Physical activity is crucial for health and disease prevention.
• Barriers exist for implementing evidence-based lifestyle recommendations.
• Artificial intelligence (AI) offers potential to augment healthcare, but its suitability for exercise recommendations is unexplored.

Purpose of the Study:

• To evaluate the comprehensiveness, accuracy, and readability of AI-generated exercise recommendations.
• Assessing a novel AI chatbot's output for clinical populations.

Main Methods:

• Developed a coding scheme for 10 categories of exercise recommendations.
• AI chatbot generated recommendations for 26 clinical populations.
• Assessed comprehensiveness, accuracy, and readability (Flesch-Kincaid).

Main Results:

• AI recommendations were 41.2% comprehensive and 90.7% accurate.
• Inaccuracies often related to preparticipation medical clearance.
• Readability was at a college level; themes included safety concerns, aerobic bias, and potential discrimination.

Conclusions:

• Significant gaps exist in AI-generated exercise recommendation comprehensiveness, accuracy, and readability.
• Exercise and healthcare professionals must recognize AI technology limitations.
• Caution is advised when using AI tools for exercise-based lifestyle changes.