Colorimetric sensor arrays and artificial intelligence for natural product quality evaluation: From cross-reactive fingerprints to analytical validation

Natural products encompass traditional Chinese medicines, herbal medicines, teas, food-derived plant materials, plant extracts, and naturally occurring bioactive compounds. Their rapid identification and quality evaluation require a balance among integrative characterization, analytical accuracy, and field accessibility. Empirical authentication methods are difficult to quantify and reproduce, whereas confirmatory techniques based on chromatography, mass spectrometry, and spectroscopy, although analytically reliable, cannot fully meet the frequent demands of on-site screening, process monitoring, and risk triage. A colorimetric sensor array (CSA) generates sample-specific color fingerprints through cross-reactive responses across multiple sensing elements. Combined with standardized image acquisition, feature extraction, and chemometric or machine-learning analysis, CSA-based workflows can support classification, quantitative prediction, and anomaly detection. This integrated strategy is referred to here as CSA-AI, an umbrella term that does not equate conventional chemometrics with artificial intelligence (AI). This review critically examines CSA design, color-data acquisition, chemometric and AI-assisted analysis, analytical validation, and applications in natural product quality evaluation along the chain of "array response-color fingerprint-computational interpretation-quality decision-making." Particular attention is given to authentic complex matrices, sample-level data partitioning, data leakage, independently sourced external evaluation, cross-device transfer, validation evidence, model interpretability, and confirmation by established analytical methods. Current direct visible-CSA evidence is strongest for authentication and preliminary adulteration screening, whereas process monitoring, storage-deterioration warning, and field deployment remain less thoroughly validated. Current evidence supports CSA-AI primarily as a complementary tool for rapid screening, risk warning, and auxiliary grading rather than as a replacement for confirmatory methods such as HPLC, GC-MS, and LC-MS.