From Benchmark Accuracy to Airworthiness-Compatible Maintenance Decision Support: AI-Enabled Aircraft Predictive Maintenance—A Critical Evidence Synthesis and Evidence-Maturity Framework

High predictive accuracy on aircraft maintenance benchmarks does not, by itself, demonstrate validity in operational settings or readiness for airworthiness-compatible decisions. This critical evidence synthesis examines prognostics, fault detection, inspection, digital twins, maintenance optimization, and governance. A curated historical bibliographic export of 764 records informed a documented relevance audit of 663 records, which was compared with an existing analytical corpus of 56 sources. Twenty additional sources expanded the corpus to 76, and 50 purposively selected empirical/technical sources supported the descriptive analysis. Under the original coding scheme, the 50 studies comprised 23 synthetic/operations-model studies (46%), 8 realistic-simulation studies (16%), 17 primarily operational-data studies (34%), and 2 cross-domain/distributed studies (4%). These mutually exclusive historical labels do not fully capture the use of real-world data. The original evidence-maturity coding assigned 13 studies to L1, 18 to L2, 16 to L3, 3 to L4, and none to L5. The author proposes separate pathways for externally demonstrated transfer (L4-T) and independently evaluated maintenance-decision validity (L4-D) and define L5 in terms of a function-specific assurance evidence dossier informed by prospective human–system assessment. The recent 2025–2026 literature and cross-sector risk frameworks are discussed narratively outside the historical study counts. The contribution is an auditable distinction between predictive model scores, operational validity, maintenance-decision value, and controlled adoption.