Credit risk is a fundamental element for banks to remain solvent. Due to financial stability considerations, central banks and regulators are particularly interested in measuring macro–micro linkages and have developed specific expertise in this area. Today, the need for machine learning (ML) and Explainable AI (XAI) is becoming more important as financial decisions made by artificial intelligence (AI) increase. XAI is needed to see the rationale behind the results of complex and opaque financial models in calculating credit risks, to minimize loss of confidence by improving risk assessments and to promote a more flexible financial system. This study highlights a number of methodological challenges that can be useful in the process of model analysis and review during the evaluation of papers applying machine learning in current financial practice. In addition, this research identifies the best ML technique and performance measure of ML techniques commonly used in credit risk prediction research. To this end, research in this area shows how various ML techniques are used in credit risk models and how these techniques affect performance. Issues such as the explainability, reliability and interpretability of the models used are important obstacles faced by ML techniques in the credit risk prediction. In addition, methodological challenges and limitations highlighted by existing studies are addressed and recommendations for future research are provided.