<p>In a recent article (‘ChatGPT is bullshit’), the authors argue that large language models (LLMs) like ChatGPT generate bullshit—a philosophical term coined by philosopher Harry Frankfurt that picks out convincing, but truth-insensitive content. The authors argue LLMs like ChatGPT are bullshitting machines. Here I defend a complementary account of why LLMs like ChatGPT produce bullshit: LLMs like ChatGPT are bullshit (partly) because people are bullshitters. Given that LLMs like ChatGPT predict and generate text based on large quantities of data from human language users, LLMs like ChatGPT would naturally reflect that fact—bullshit in, bullshit out. This paper then explores the incentives for people to bullshit—to manage reputations, to signal their intelligence, and to participate in the marketplace for bullshit—along with why such incentives will likely motivate people to use LLMs like ChatGPT to better bullshit each other.</p>