During a survey of an experiment that involved using "carbon calculators", we investigated what people think and what they do when they consider the link between their food and the climate. To what extent do individuals adhere to or reject the "prescriptions" implicitly contained in carbon calculators and passed on by environmental social movements or the authorities. These prescriptions are considered as recommendations or advice drawn up by experts aiming to guide behaviours; they involve reducing reality in a measuring device against which behaviours are gauged (in this case, the quantity of greenhouse gases). Three main levers are envisaged so that households can reduce their food's impact on the climate: reducing the consumption of meat, eating local and eating in-season produce. However, in opposition to this conception of individuals as rational and free to adapt their practices based on the information available to them, we suggest an analytic framework that takes into consideration the fact that food and eating is embedded in everyday and social relationships. Calculating one's greenhouse gas emissions (GGE) provokes an intense moment of reflexivity during which routines are called into question and discrepancies between prescriptions and practices emerge. Drawing our inspiration from, on the one hand, the theory of practice, we examine how the prescriptions are connected to practices, and on the other, using a pragmatic approach, we analyse how individuals call into question their food practices to resolve the contradictions they identify, and devise adjustments. We show how people appraise the context globally, which leads them to subvert the logic of the calculator or address the prescriptions more broadly to lend value to other practices to which they attach importance and consider adapted. This enabled us to highlight vegetable gardening and promoting eating locally produced food as favoured adjustments. The qualitative approach allows us to emphasise how the measuring device, rather than supplying a behaviourist model, makes it possible to render more visible the forms of attachment and categories of judgement mobilised, which do not obey strict accounting logics or a formal, narrow rationality, and to examine "what counts when you're counting".