The number of areas in which artificial intelligence (AI) technology is being employed increases continually, and climate change is no exception. There are already growing efforts to encourage people to engage more actively in sustainable environmental behavior, so-called “green nudging”. Nudging in general is a widespread policymaking tool designed to influence people’s behavior while preserving their freedom of choice. Given the enormous challenges humanity is facing in fighting climate change, the question naturally arises: Why not combine the power of AI and the effectiveness of nudging to get people to behave in more climate-friendly ways? However, nudging has been highly controversial from the very beginning because critics fear it undermines autonomy and democracy. In this article I investigate the ethics of AI-powered climate nudging and address the question whether implementing corresponding policies may represent hidden and unacceptable costs of AI in the form of a substantive damage to autonomy and democracy. I will argue that, although there are perfectly legitimate concerns and objections against certain forms of nudging, AI-powered climate nudging can be ethically permissible under certain conditions, namely if the nudging practice takes the form of what I will call “self-governance”.