When the guess becomes a default: Imagined audiences as production constraints

Taking up a question opened by Siao Yuong Fong’s recent Cultural Commons essay, ‘Imagining audiences in the age of generative AI’, this essay asks how audience imaginaries acquire operational force inside production and take aesthetic form. Fong’s palimpsest foregrounds how heterogeneous audience assumptions accumulate; the default names which of them returns with the least resistance and which becomes costly to refuse. Drawing on an AI-enhanced virtual-production archive in South Korea, and on a history running from radio’s absent listener to the mechanical storage of audience response in 1950s television, the essay argues that generative AI does not make actual audiences more knowable. It makes the audience position sedimented in the apparatus partially reconstructible – through recurrent outputs and the labour required to undo them. Model defaults meet budgets, workflows and cultural expertise, leaving formal traces in framing, movement, atmosphere and signage. Following those traces shows who pays when cultural specificity does not arrive by default.