Background?Methamphetamine (MA) is one of the most commonly abused illicit psychostimulant drugs and MA use disorder constitutes a universal health concern across the world. Despite many intervention approaches to MA use disorder, the indicator of addiction severity is mainly limited to subjective craving score to drug-related cues, which is influenced by many factors such as social approval and self-masking. Aim?The present study investigates whether self-reported craving for drug use in response to MA cues is a reliable indicator for addiction severity in MA users, and then tests the validity of the cue-induced attention bias test in addiction severity assessment. Methods?Fifty-two male MA users completed the cue-induced craving test and attention bias task, and were required to report clinical characteristics of addiction severity. For the attention bias test, subjects were required to discriminate the letter superimposed onto MA use-related or neutral scenes. The reaction time delay during MA-use condition relative to neutral condition was used as an index of the attention bias. Results?The results showed that 24 of the 52 MA users rated non-zero in cue-induced craving test, and they showed a significant attention bias to drug-related pictures. However, the other 28 users who rated zero in cue-induced craving evaluation showed a similar attention bias to drug-related cues. In addition, the attention bias to MA use-related cues was significantly and positively correlated with the clinical indexes of addiction severity, but the relationship was absent between subjective craving evaluation and the indexes of addiction severity. Conclusion?These results suggest that attention bias to MA cues may be a more reliable indicator than experiential craving report, especially when subjective craving is measured in the compulsory rehabilitation centre.