Cancer is a major public health problem worldwide. Gastrointestinal cancers account for approximately one-third of the total global cancer incidence and mortality. Historically, the mechanisms of tumour initiation and progression in the gastrointestinal tract have been studied using cancer cell lines in vitro and animal models. Traditional cell culture methods are associated with a strong selection of aberrant genomic variants that no longer reflect the original tumours in terms of their (metastatic) behaviour or response to therapy. Organoid technology has emerged as a powerful alternative method for culturing gastrointestinal tumours and the corresponding normal tissues in a manner that preserves their genetic, phenotypic and behavioural traits. Importantly, accumulating evidence suggests that organoid cultures have great value in predicting the outcome of therapy in individual patients. Herein, we review the current literature on organoid models of the most common gastrointestinal cancers, including colorectal cancer, gastric cancer, oesophageal cancer, liver cancer and pancreatic cancer, and their value in modelling tumour initiation, metastatic progression and therapy response. We also explore the limitations of current organoid models and discuss how they could be improved to maximally benefit basic and translational research in the future, especially in the fields of drug discovery and personalized medicine.