The Last Costly Signal: How Generative AI Collapses Competence Signaling and Why Liability Sustains Markets for Expert Services

Abstract

Generative artificial intelligence has driven the cost of producing convincing artifacts of expertise toward zero. Signaling theory predicts that signals whose informational content rests on production cost lose that content when production becomes cheap. We formalize this for markets for expert services, a class of credence goods, modeling AI as a compression of the discernible headroom between what machines produce at negligible cost and what buyers can distinguish. Below a critical headroom no separating equilibrium in artifact-based production-side signals exists—for any single-crossing cost family, with the buyers’ discernment ceiling held fixed: the market pools, the competence premium vanishes, and able providers exit. We show that an outcome-contingent signal—a warranty backed by damages D with ex-post verifiability φ—sustains a fully separating equilibrium at any level of AI capability whenever φD≥v (strictly for φD>v), where v is the value of a solved problem, under four institutional preconditions stated explicitly and priced in turn. The expected cost of liability turns on whether the problem is solved, not on how cheaply documents are produced, and is invariant to AI capability. Provenance certification priced as a type-independent stamp cannot restore full separation; a verified commitment to produce without generative assistance can, at the pre-AI signaling cost. Two further results endogenize the contract’s institutions: liability signaling has a minimum ticket size set by the fixed costs of a civil procedure and, under insurance, the signal-effective quantity is the retained, collectible exposure.