Comparison of Web‐Based Image Identification Platforms and DNA Barcoding for Identification of Mature Weed Plants

ABSTRACT Accurate species identification is essential for effective weed management. However, conventional morphology‐based approaches are limited by expertise shortages and ambiguous diagnostic characters. This study evaluated the performance of three web‐based image identification platforms (iNaturalist, PlantNet, Google Lens) and DNA barcoding ( rbc L, mat K) against morphological identification for 39 random mature weed plants from Jelebu, Malaysia. Preliminary morphological identification resolved 71.8% of the plants to the species level. PlantNet, iNaturalist, and Google Lens identified 92.3%, 84.6%, and 84.6% of the plants to the species level, respectively. All three platforms produced concordant identifications for 71.8% of plants and confirmed the preliminary identifications in 61.5% of cases. By contrast, rbc L yielded species‐level matches for all plants, whereas mat K failed to resolve two plants to the species level. Marker discordance between rbc L and mat K was observed in 10 plants. Comparison between rbc L and identification applications showed only 41.0% matches across all platforms, whereas 46.2% were wholly discordant. Notable reassignments by barcodes illustrate visually plausible but incorrect morphological identifications. Our results indicate that image identification platforms offer rapid, user‐friendly tools but can produce substantial misidentifications. Molecular barcodes provide higher objectivity but have limitations like accessibility, marker discordance and database issues. We recommend an integrated workflow between image‐based platforms, targeted barcoding for low‐confidence cases, and expert validation to improve identification accuracy for operational weed management and broader biodiversity applications.