The social composition of online spaces such as social media platforms has undergone a recent dramatic transformation with the advent of generative AI. Online material generated by AI includes human-like chatbot interactions and manufactured text and images purporting to represent real-life people, places and events. Efforts to estimate the extent of such material are hampered by the considerable difficulty of reliably detecting machine-generated material. For some commentators, these developments threaten the ability of online spaces to offer meaningful social engagement. This paper explores whether, in this context, it is still possible to conduct online ethnography aimed at understanding culturally-embedded meaning-making. The paper argues against a generalized methodological exceptionalism for generative AI. Instead, some promising strategies are found in existing methodological approaches that treat authenticity as a problem experienced by ethnographers and participants. A reflexive approach to ethnographic treatment of authenticity remains a valuable stance in situations where suspicions about the presence of generative AI are raised. In particular, multi-sited approaches allow experience of varying and cross-contextual understandings of authenticity and autoethnography focuses attention on how we navigate the lived experience of uncertainty about the nature of online content. Second, the paper turns to more-than-human ethnographic approaches and finds ethnography positioned as an immersive means to embrace non-human actors, including AI-generated features, as an intrinsic part of online experience. Such approaches ask for reflexivity around what is at stake in making judgments about the ontological state of materials encountered online. The methodological strategies reviewed here suggest that there is a future for online ethnography in the face of generative AI involving ongoing methodological innovation without wholesale methodological exceptionalism, but that this requires both a multi-faceted reflexivity and caution in adopting human-centric approaches founded on principled separability of human and machine.