Validating ChatGPT output against the scientific literature