Validation, Calibration and Transparency, More than Accuracy: An Appraisal of Artificial Intelligence Against Antimicrobial Resistance and a Sustainable, Testable Diagnostic Architecture

Artificial intelligence is widely proposed against antimicrobial resistance, but how much of that literature could support clinical deployment is unclear. We searched PubMed (2016–2026) across four domains, retrieving all 2446 records. Of 725 eligible studies, 80 were appraised at full text against the artificial-intelligence appraisal extensions (PROBAST+AI, TRIPOD+AI). Selection favoured externally validated work, so the validation figures are upper bounds. Output grew from 10 studies in 2016 to 142 in 2025; within the appraised set there was no evidence that genuinely external evaluation became more likely (odds ratio 1.08/year, 95% CI 0.89–1.34). Of the 80, 45% were externally validated, 24% released a model a reader could apply, 23% assessed calibration, 16% justified a sample size, and five compared performances across patient characteristics against eight across sites; 8.8% called a random split external validation. Risk of bias was low in 16, unclear in 55 and high in nine, and none of the 18 deep-learning studies reported calibration (95% CI 0–17.6%, a loose bound on 18). We derive the design constraints these impose on a culture-independent diagnostic platform, a projection with no performance data. Validation, calibration and transparency limit this field more than accuracy does: discrimination was reported in every study appraised.