The evolution of intelligence is one of nature’s greatest mysteries. We have little idea how we came to be so clever, or even how our minds actually differ from those of other species. For over 100 years researchers have compared human and animal intelligence in order to gain insight into this area. However, these comparisons have been based on a “success-testing” approach, where researchers simply examine whether an animal can solve a problem or not. This approach has generated a great deal of debate, but little progress, because success-testing is flawed. Just because an animal solves a problem, it does not follow that the animal has used the same cognitive mechanisms as humans to do so. Inspired by Alan Turing’s work on machine intelligence, I will argue we need to adopt a novel “signature-testing” approach to avoid this problem. By searching for the signatures of cognitive mechanisms (their limits, errors and biases) it is possible to demonstrate whether a human and an animal truly think in the same way or not. I will outline how signature-testing can chart the evolution of intelligence and demonstrate its potential for two areas of comparative cognition: causal reasoning and metacognition.