AI-Based Ayurvedic Treatment Planning Systems

Abstract

Artificial intelligence systems for Ayurvedic treatment planning have moved fairly quickly from research prototypes to deployable clinical tools over the past decade, but the evaluation landscape has not kept pace. Published systems cover constitutional classification, herbal recommendation, pulse analysis and integrated decision support, yet each is typically reported against its own idiosyncratic benchmarks, which makes cross-system comparison frustrating. We pulled together 64 peer-reviewed AI-for-Ayurveda systems described between 2012 and 2024, and sorted them into five system frames: A1 (Prakriti classification), A2 (herbal-formulation recommendation engines), A3 (Nadi Pariksha pulse-analysis systems), A4 (classical-text NLP and knowledge extraction), and A5 (integrated clinical decision support). Each frame was scored with the Ayurvedic AI System Validation Index (AAVI), a weighted composite across classification accuracy, clinical utility, practitioner agreement, system robustness, and deployment evidence. Prakriti classification systems (A1) and integrated clinical decision support (A5) came out on top at AAVI 0.882 and 0.848, with the former carried by strong accuracy numbers against practitioner gold-standard labels. Text-NLP systems (A4) scored well on extraction benchmarks but less well on clinical utility. Our reading is that the field is now technically mature for classification but still under-developed on the integration side, where the real clinical value is likely to be.