Clinical readiness of artificial intelligence for treat-to-target management in inflammatory bowel disease: from scores to decisions

Artificial intelligence (AI) is increasingly used in inflammatory bowel disease (IBD), particularly for endoscopy, histology, transmural imaging, and prediction of treatment response. However, strong diagnostic or predictive performance does not necessarily confer clinical utility. This narrative review traces the development of AI in IBD from objective disease assessment and outcome prediction to therapeutic decision support within a treat-to-target framework. Endoscopic AI has the most mature evidence base, while applications in histology and transmural imaging are at earlier stages of clinical translation. Models of treatment response can stratify patients by the likelihood of response, remission, relapse, or treatment sustainability; most, however, remain prognostic rather than prescriptive. We therefore propose the clinical readiness of AI in IBD framework, comprising analytical validity, clinical validity, generalizability, workflow utility, decision utility, and clinical impact. The strongest evidence concerns disease measurement and prediction; evidence that AI-informed or AI-guided strategies improve treatment decisions and patient outcomes remains limited. Future research should prioritize external validation, prospective evaluation in clinical workflows, comparative trials of prespecified AI-informed or AI-guided treatment strategies, and patient-important outcomes.