Exercise addiction (EA) refers to excessive exercise, lack of control, and health risks. The Exercise Addiction Inventory (EAI) is one of the most widely used tools in its assessment. However, the cross-cultural psychometric properties of the EAI could be improved because it misses three pathological patterns, including guilt, exercise despite injury, and experienced harm. Therefore, the present study tested the psychometric properties of the expanded EAI (EAI-3) in a large international sample. The EAI-3 was administered to 1931 physically active adult exercisers speaking five languages (Chinese, German, Italian, Japanese, and Turkish) and other measures for obsessive–compulsive behavior, eating disorders, and personality traits. The assessment structure and reliability of the EAI-3 were tested with factorial analyses and through measurement invariance across languages and sex. Finally, a cutoff point for dysfunction-proneness was calculated. The EAI-3 comprised two factors, reflecting the positive and pathological sides of exercise. The structure had excellent reliability and goodness-of-fit indices and configural and metric invariances of the scale were supported. However, three items caused violations in scalar invariance. The results of partial measurement invariance testing suggested an adequate fit for the data. Following sensitivity and specificity analysis, the EAI-3’s cutoff score was 34 out of a maximum score of 48. This preliminary study suggests that the EAI-3 is a promising tool for screening EA in an international sample, with a robust and reliable structure comparable across languages and sex. In addition, the proposed cutoff could pave the way toward a consensus on a threshold to screen for EA.