The diagnosis of drug-induced liver injury is often one of exclusion with initial suspicion based on circumstantial evidence. The natural history, characteristics and limitations of this exclusion process are revised. Also, the numerous published attribution algorithms for evaluation of drug-related liver abnormalities are described and their characteristics and differences are illustrated with true patients from our clinical experience. Situations that complicate the diagnosis such as age, sex, concomitant use of other drugs, genetic polymorphism in metabolic pathways involved in activation or disposition of therapeutic drugs and drug-drug interactions are described. Finally, developing approach to diagnosis of drug-induced liver injury, different of attribution algorithms, are evaluated and explained using a new method based in a Bayesian approach developed and published by the authors. The authors' vision of all these potential advances and their clinical utility is provided.