Synthetic data, generated through advanced artificial intelligence models, are gaining traction in healthcare research, particularly in high-stakes fields such as haematology and oncology. By replicating statistical properties, intervariable relationships and behaviours of real-world data, synthetic data sets can serve as valuable supplements or substitutes for conventional medical data. They offer the potential to overcome barriers to data access and sharing, democratize scientific discovery, and reduce the costs and failure rates of clinical trials. However, the lack of standardization in training data selection, model evaluation, bias mitigation, privacy preservation and quality assurance remain major challenges, limiting their reliability and safe application. In this Review, we explore the role of synthetic data in cancer research and clinical trials, present real-world examples of their use, critically examine limitations and pitfalls, and propose best practices to enhance fidelity, validity, fairness and utility. Although synthetic data are not a 'silver bullet' for the challenges of clinical research, with rigorous validation and oversight, they have the potential to transform data sharing, scientific collaboration and clinical trial design.