Nondestructive evaluation (NDE) is moving from manually interpreted indications toward connected measurement systems in which sensing, imaging, automated processing, artificial intelligence (AI), and engineering decisions are coupled through a digital thread. This transition increases information availability but also introduces additional points at which measurement meaning, uncertainty, or traceability can be lost. This critical integrative review reframes NDE 4.0 as a metrological-assurance problem and proposes a six-layer architecture spanning the physical measurand, instrument, signal, information, inference, and decision layers. The architecture is combined with an uncertainty map for hybrid physical-digital inspection, a minimum digital evidence record, and a qualification V-model linking intended use to verification, representative validation, bounded autonomy, and post-deployment surveillance. A worked ultrasonic-sizing example illustrates how classical measurement contributions and model-inference uncertainty can be combined without treating AI confidence as a substitute for measurement uncertainty. The synthesis distinguishes metrological traceability from digital provenance, probability of detection from classification metrics, and probabilistic calibration from measurement uncertainty. Cross-sector analysis of aerospace, nuclear, rail, civil infrastructure, and additive manufacturing shows that increasing autonomy raises both information value and evidential burden. The central conclusion is that, for measurement-based engineering decisions, especially in higher-consequence applications, the defensibility of an NDE 4.0 system depends on an adequately traceable, uncertainty-aware, intended-use validation, and auditable sensor-to-decision chain.