AI-Driven Personalization of Ayurvedic Prescriptions

Abstract

Ayurvedic prescription is, in principle, a personalised clinical decision. Classical practice derives each prescription from the patient's prakriti (constitutional type), the vikriti (current imbalance state), the dominant symptom pattern, concurrent conditions, and an assessment of digestive and metabolic fire (agni). The same chief complaint produces genuinely different prescriptions for different patients. That personalisation principle is also what makes Ayurvedic clinical research hard to generalise: the treatment arm in an Ayurvedic trial is often not a single intervention but a family of individualised interventions. AI-driven personalisation approaches attempt to systematise this prescribing work in ways that are both reproducible and respectful of classical personalisation logic. This study reviews 62 primary studies published between 2010 and 2021 applying AI methodology to Ayurvedic prescription personalisation, grouped under five approach frames: P1 (rule-based personalisation engines with encoded classical knowledge), P2 (supervised ML recommender systems trained on practitioner prescription data), P3 (collaborative filtering and similarity-based systems), P4 (knowledge graphs combined with machine learning), and P5 (multi-modal deep learning personalisation). Each frame was scored using the AI Personalisation in Ayurveda Index (APII). Knowledge graph + hybrid systems (P4) scored highest at APII 0.842, with rule-based engines (P1) a strong second at 0.802 driven by their interpretability advantage.