Akaike Information Criterion (AIC) is a statistical measure used to compare models, balancing goodness of fit and model complexity by penalizing the number of parameters. It helps in model selection by choosing the model with the lowest AIC value, indicating the best trade-off between fit and simplicity.