Has the world gone botshit crazy? A response to the Frankfurtian critique of ChatGPT in higher education

Abstract Hicks et al. (Ethics Inf Technol 26:38, https://doi.org/10.1007/s10676-024-09775-5 , 2024) argue that ChatGPT is indifferent to truth and produces bullshit in the Frankfurtian sense Frankfurt (On Bullshit, Princeton, 2005). Warning against hype around these tools in education, they offer supporting semantic and pragmatic arguments. This article shows how existing responses in Fisher (Ethics Inf Technol 26:67, https://doi.org/10.1007/s10676-024-09802-5 , 2024), Tigard (AI Ethics, https://doi.org/10.1007/s43681-025-00743-3 , 2025), and Gunkel and Coghlan (Ethics Inf Technol 27:23, https://doi.org/10.1007/s10676-025-09828-3 , 2025) fail to address the argument on its own terms or only partially address it. Clarifying their case, I offer a novel rebuttal that addresses their framework fully and on its own terms. I provide responses to both the semantic and pragmatic strands of their case. Semantically, I show that the Frankfurtian critique of ChatGPT has implausible implications and is incoherent in its broader aims. However, the reasons for its failure provide important insights into the functional nature of Large Language Models. Drawing on Wright (Philos Rev 82:2, 1973), and Houkes and Vermaas (Technical functions: on the use and design of artefacts. Springer, Dordrecht, 2010) I distinguish between several different philosophical models of functionality and offer a novel analogy to better understand the nature of LLMs. Pragmatically, I present evidence that the rhetoric of bullshit is counterproductive in the educational field, a key policy arena. I set out a new typology of approaches to GenAI in education and describe how the bullshit-frame is inconsistent with current pedagogical consensus on best practice. I show how this rhetoric undermines key elements in emerging harm reduction strategies, and obscures potential opportunities in both the technology and the educational reforms that face us. This casts doubt on the value of Frankfurtian-style critiques but leaves the door open to better targeted AI-critical concepts.