We assess the communications of 37 airlines on their own websites regarding voluntary carbon offsets (VCO) to determine the extent to which they are either trustworthy or misleading. We propose an innovative coding framework that captures the trustworthy or misleading attributes of the messages as they are applied to: i) the type of claim (product, process, fact or image), and ii) the nature of the claim (fibbing, hidden trade-off, no proof, vagueness, irrelevance, lesser of two evils or worshiping false labels). We deploy a quantitative, multi-method approach that combines content analysis and discrete choice modelling, and we corroborate the taxonomy developed with lexical analysis. We identify the various factors that affect the pattern of 56% of claims being trustworthy and 44% being misleading. We demonstrate how a combined study of the trustworthy or misleading characteristics of communications provides more learning opportunities than studying either individually.