Toward AI-Ready Paper-Based Colorimetric Glucose Sensing: Selectivity, Sensitivity, and Standardized High-Throughput Data Generation

Paper-based colorimetric glucose sensing has evolved from simple visual assays toward digitally quantified systems increasingly supported by artificial intelligence (AI). However, advanced computational analysis cannot fully compensate for limitations in sensing hardware or for insufficient and poorly standardized analytical data. In this Perspective, we argue that truly AI-ready paper-based glucose sensing requires coordinated development of three capabilities: selective molecular input, sensitive and reproducible signal generation, and standardized high-throughput data generation. Selective input can be engineered through filtering and permselective interfaces; signal generation can be strengthened through catalytic amplification, enzyme stabilization, reagent organization, and reaction-microenvironment engineering; and standardized paper platforms can provide the large, comparable datasets required for robust machine-learning models. We discuss these developments in the context of paper-based glucose assays and propose an integrated framework in which sensing chemistry, device architecture, dataset design, and computational analysis are co-designed. Such integration shifts AI from compensating for limitations of individual assays toward extracting robust quantitative information from intrinsically selective, informative, and reproducible analytical data.