Generative AI threatens knowledge online not merely by producing falsehoods, but by changing the epistemic environment in which online beliefs are formed. I argue that the spread of convincing AI-generated counterfeits creates a local skeptical challenge to digital testimony. The challenge is not that ordinary subjects must rule out every skeptical possibility. Rather, for many beliefs formed solely from unauthenticated online materials, ordinary subjects are situated among nearby, method-indist…
Read moreGenerative AI threatens knowledge online not merely by producing falsehoods, but by changing the epistemic environment in which online beliefs are formed. I argue that the spread of convincing AI-generated counterfeits creates a local skeptical challenge to digital testimony. The challenge is not that ordinary subjects must rule out every skeptical possibility. Rather, for many beliefs formed solely from unauthenticated online materials, ordinary subjects are situated among nearby, method-indistinguishable bad cases. In such environments, even true beliefs formed from genuine digital items may fail to satisfy a familiar anti-luck condition on knowledge. This does not show that we never know from online sources. It shows instead that digital-only uptake often falls short of knowledge unless supported by provenance, corroboration, or other independent checks. I close by suggesting that this narrowing of online knowledge may reinforce existing patterns of epistemic injustice.
