The ultimate forgery: A semiotic deconstruction of large language models as engines of simulated selfhood

Abstract Information science rests on a compact: its mediating signs maintain a traceable relation to what they describe. Large language models (LLMs) break it architecturally, generating discourse that is statistically plausible yet epistemically ungrounded, while standard metrics (precision, recall, relevance) cannot detect the rupture. The field therefore lacks not a further metric but a theory of signs that appear authoritative while ungrounded; we propose critical semiotics as that theory. Treating outputs as observable sign‐actions, we argue the LLM is an engine of the ultimate forgery: a simulation whose object is meaning‐making itself. A stratified reading (Saussurean langue, Derridean écriture, Peircean process semiotics, sharpened by Leone's “absolute fake”) yields Structural Selfhood: an enunciative “I” sustained by recursive architecture rather than interiority. The rupture is located precisely: the model satisfies Peircean speculative grammar and functions methodeutically, while critic , where a sign is held answerable to its object, has no locus in the architecture. Two consequences follow: a systemic erosion of veridiction, that is, of the capacity to distinguish truth‐telling from its simulation, and generative algorithmic epistemic injustice, the presentation of a corpus's distribution as knowledge itself. Mata v. Avianca's fabricated citations render the mechanism legible. Critical semiotics is thus an indispensable analytical resource.