A fundamental challenge in neuroscience is to understand how the brain processes information to make sense of the environment. How do we recognise that an image shows London instead of Glasgow? Neuroscientists often approach this question by measuring brain activity at different granularities. However, we must go beyond this and understand what these activities represent in terms of the specific stimulus information processed. We make this brain-activity-to-information leap with new tools that can reconstruct from brain activity the information processing that leads to recognition. Our novel methodology precisely varies stimuli to modulate the brain and behavioural responses of humans who recognize them (e.g. as “London” or “Glasgow”). From these variations, we can reconstruct the information processing networks that reveal where, when and how the brain processes specific stimulus-information to flexibly categorise scenes, objects and faces to behave adaptively.