Artificial intelligence in clinical gastroenterology is expanding diagnostic tools and workflows

A special issue of The American Journal of Gastroenterology reviews current studies on AI across gastroenterology, reporting mixed results for polyp detection, faster AI‑assisted capsule endoscopy review, LLM‑based decision tools, and discussions of ethical and legal concerns.

A special issue of The American Journal of Gastroenterology gathers recent studies and commentary on artificial intelligence (AI) in gastroenterology. The collection examines computer vision for endoscopy, large language models (LLMs) for clinical reasoning and recommendations, automated workflows, and early robotics research. Authors report mixed performance across tasks and highlight practical, ethical, and legal implications as AI tools become more accessible to clinicians. The editorial notes that clinicians will need to assess AI value and performance as use expands. Key findings: - Studies of computer‑aided detection (CAD) in colonoscopy showed mixed results: a Mayo Clinic real‑world series in 4,000 patients reported improved polyp and adenoma detection, while a 1,600‑subject multicenter German study found no difference; a Canadian report described "alert fatigue" linked to reduced adenoma detection over a full day of screening colonoscopy. - An international prospective real‑world study reported AI‑assisted video capsule endoscopy could review complete studies in about four minutes and outperformed standard gastroenterologist interpretation in that setting. - LLMs were reported to perform automated colonoscopy surveillance recommendations with strong concordance to guideline follow‑up, and one report noted modern LLMs can pass GI board examinations. - Other reported applications include an FDA‑approved AI colonoscopy quality assessment tool, AI models for anorectal manometry interpretation, hepatology tasks such as summarizing records and identifying encephalopathy or decompensated cirrhosis, patient coaching (e.g., stool image guidance and diet advice), and early work on magnetically driven robotic colonoscopy. - Authors and commentaries emphasize concerns about accuracy, clinical impact, workflow effects, provider deskilling, medico‑legal responsibilities, privacy, and economic implications. Summary: The assembled studies suggest AI is producing task‑specific gains in gastroenterology while yielding mixed results in other areas. Reported benefits include faster review times, automated recommendations, and early robotics research, while uncertainties remain about diagnostic value, alert fatigue, and broader clinical impact. Regulatory, ethical, and legal discussions are proceeding alongside clinical evaluation. Undetermined at this time.