-
AI Language Models Show Promise in Digestive Disease Care, But Evidence Remains
Artificial Intelligence (AI) language models such as ChatGPT are showing early potential in improving patient education and trust in digestive disease care, but strong evidence from real-world patient outcomes is still limited, according to a new review published in Gastroenterology & Endoscopy.
The scoping review analysed 14 randomized controlled trials (RCTs) assessing the use of large language models (LLMs) in diagnosing and managing digestive disorders. However, only four studies involved real patients, highlighting a significant evidence gap.
Digestive diseases affect billions globally and often require complex diagnostic and treatment pathways. While AI could help streamline decision-making, enhance communication, and reduce administrative burden, concerns remain around accuracy, bias, data privacy, and reliability of AI-generated outputs.
The review found that LLMs are currently used mainly for patient education and clinical decision support, rather than direct treatment outcomes. Trials showed improvements in patient understanding, trust, and acceptance of AI-assisted care, but limited proof of improved clinical outcomes.
Domain-specific AI tools, such as applications designed for colorectal cancer screening or postoperative symptom monitoring, were found to perform better than general-purpose models in some tasks. Most trials were conducted in China and the United States, often at single centres with small sample sizes.
Researchers emphasised that while early results are encouraging, larger, well-designed RCTs using real patient data are essential to determine the true clinical value of AI in digestive healthcare.
