Artificial intelligence (AI) is increasingly used in sports for movement analysis, performance monitoring, injury research, talent identification, officiating support, and fan services [23]. This review assesses how well the current evidence supports these claims and what uncertainties remain. The strongest evidence is with respect to specific measurement tasks. For instance, a validation study of markerless pose estimation in 5 participants reported mean absolute joint angle errors of 9.7 ± 4.7 degrees for athletic movements and 9.0 ± 3.3 degrees for sports movements [1]. A separate study found that VAR improved decision accuracy from 92.1% to 98.3% across 2,195 matches [2] (again, this refers to video-assisted refereeing, not AI). Prediction-based applications are less settled. Talent-identification studies report ROC-AUC values from about 0.58 to 0.96, but calibration is rarely reported, external validation is uncommon, and 59% of studies were judged at high risk of bias [3]. A return-to-sport review also found that most included studies were rated unclear or high risk in the analysis domain, primarily because external validation remained absent [4]. The data suggest that AI can serve as a measure or decision-support tool, provided it's validated for the specific application. There are no apparent broad predictive or clinical benefits. The next step is to validate this independently, report it properly, involve the practitioner, and set clear rules for using athlete data. This is a narrative review rather than a systematic review, and the sources we've used are not exhaustive. They should not be interpreted as a full representation of the literature that's available