Fans are increasingly aware of deepfake—believable AI-fabricated videos—and are therefore more skeptical of unverified information, even when visual evidence appears convincing. This article offers a methodological reflection on analyzing a deepfake event in which fans produced and circulated AI-generated disinformation to playfully undermine the credibility of a celebrity’s video scandal. We explore the complex human-community-machine interactions (HCMI) between fans and AI-generated images, and we discuss how researchers can ethically (re)present their findings. We call for rethinking the “fan first” principle, a core tenet of ethical fandom research. Drawing on Puig de la Bellacasa’s technoscientific theorization of care, we propose a critically speculative ethics of care in fandom research, guided by three principles: (1) thinking with fans, (2) thinking for fandom, and (3) thinking beyond fans and fandom. This approach is particularly relevant in a digital media ecology where generative AI and fan practices mutually transform each other. Our discussion also serves as a springboard for further explorations of ethics related to AI, including its impact on trust, social relations, and power in fandom research.