Ecology has been characterized as a collection of case studies, but if results cannot extend beyond a single study, they are not really useable and maybe not really science. Can ecological data and studies ever be generalized? FAIR standards and methodology may help to reveal where heterogeneity cannot be resolved and alternatively, where synthesis is possible. It will not help where data don’t exist, are not in retrievable form, or where meta-data fails and key words are insufficient. But machine learning and interoperability are dangerous tools, and these methods can also be used to endanger privacy and even democracy.