Why does AI discourse feel obscene in Baudrillard's sense?

Baudrillard uses "obscene" in a specific technical sense — not moral, but spatial. The obscene is what appears outside the scene, stripped of the mise en scène that would make it legible as a move within a context. Pornography is obscene in this sense not because of sex but because it exposes the act outside the scene of intimacy and relation. The obscene is total visibility without situation.

AI-generated arguments fit this definition precisely. They are informationally adjacent to the cultural discourse they reference — they contain the right vocabulary, the right citations, the right argumentative form. But they are disembedded from the scene of argument itself: the relation between speakers, the history of prior exchanges, the social stakes that make the claim a move worth making. They are fully visible and fully detached.

This explains a persistent discomfort readers report about AI text that cannot be reduced to factual errors or style: the text is doing something that looks like argument, while being disconnected from the scene in which arguments are staged. It is argument-information without argument-relation. The Baudrillardian frame names this precisely — it is the obscene version of discourse, not the false version.

Production-without-process — the too-visible dimension. The obscenity claim extends beyond scenic displacement into the production side. AI knowledge is too-visible: comprehensive, immediate, total, with nothing hidden behind it — because there is nothing behind it. The analysis without the analyzing, the expertise without the experience, the conclusion without the journey. Where expert discourse is staged partly because the production process leaves marks (hesitation, revision, acknowledged uncertainty, the visible evidence of someone working through a problem), AI output presents only the finished product without the production that would normally mark its path. The ob-scene of AI knowledge is thus double: it is off-stage both in the sense of missing the relational scene of argument and in the sense of missing the productive scene of inquiry. The finished surface is continuous — no seams, no traces of effort — and that continuity is precisely what exposes its disembedding.

Hyperreality as epistemic condition. Pushed one step further, this is not merely disembedded argument but hyperreality in the operational sense: AI produces simulations of expertise that precede the expertise itself. The expert analysis exists before any expert analyzed anything. The map is generated without reference to a territory — not as cultural commentary on postmodern signification, but as a concrete epistemic condition produced by the technology. The simulation of expert speech becomes the first form in which expertise appears in the discourse, and the territory (the genuine understanding that expert speech would normally track) is retroactive at best, absent at worst. This is hyperreality not as metaphor but as epistemic specification.

The implication: AI's epistemological problem is not primarily about truth or accuracy. It is about scenic displacement. Fixing hallucination would not fix this, because the problem is structural rather than propositional. How does AI writing escape the conversations that govern knowledge? describes the same dislocation from another angle.