Disclosed are systems, methods, and non-transitory computer-readable media for differential bid generation using machine learning. A bid management system generates an input based on features of a content item and. The bid management system uses the input to determine acceptance probability score distributions for a set of candidate users. The bid management system uses the acceptance probability score distributions along with user attributes describing the set of candidate users along with campaign parameters to determine a subset of the candidate users to receive an offer to view the content item along with bid values to offer to each candidate user.