Conclusions

AI-enabled ECG represents a promising but still developing approach for potassium assessment. Current evidence was strongest for hyperkalemia detection, whereas hypokalemia detection, potassium-status classification, and continuous potassium estimation remain comparatively less mature. Although many studies reported good diagnostic discrimination, the evidence was dominated by retrospective development and internal-validation designs, with relatively few studies progressing to prospective or clinical-utility evaluation. AI-ECG should therefore be regarded as an adjunctive tool to support triage and monitoring rather than a replacement for laboratory potassium measurement. Future research should prioritize prospective multicentre validation, standardized reporting, calibration, reproducibility, and evaluation of patient-centered and workflow outcomes to establish clinical effectiveness.