Document Type

Article

Publication Title

MRS Journal of Arts, Humanities and Literature

Abstract

This article argues that AI prompting should be understood not as a tool, technique, or accessory to image generation, but as a medium in its own right. Drawing on Clement Greenberg’s theory of medium specificity, it contends that a medium becomes legible through the operations peculiar to itself, namely, the constraints, affordances, and formal procedures that distinguish it from neighboring arts. Existing scholarship has asked whether prompting counts as art, whether prompt engineering constitutes a creative skill, and whether the prompt itself can be aesthetic; however, these accounts often stop short of a sustained theory of prompting as medium. This essay addresses that gap by identifying the medium-specific property of AI prompting as an iterative exchange structured by probabilistic unpredictability. Unlike painting, sculpture, or filmmaking, where resistance may be material, embodied, and progressively masterable, prompting confronts the artist with outputs generated through opaque statistical processes that cannot be fully anticipated or reasoned with. The prompter therefore works through recursive reformulation, selection, and response, shaping the work by negotiating deviation rather than executing intention in a linear fashion. This generative interface, rather than any single prompt or isolated output, constitutes the core artistic practice of the medium. The article further argues that objections concerning prompt shareability or the instability of AI authorship mistake the art object for the process.

Research Highlights

  • The Problem: Existing scholarship fails to provide a sustained, rigorous art theory of AI prompting as a medium, instead treating it restrictively as a learned skill, a tool, or a derivative technical competence attached to a machine.

  • The Method: The researchers apply Clement Greenberg's modernist theory of medium specificity, alongside media frameworks by Niklas Luhmann and Piotr Krajewski, to analyze the constraints, affordances, and generative interface of text-to-image AI systems.

  • Qualitative Finding: AI prompting qualifies as a distinct, sui generis artistic medium because its defining operations consist of a recursive, processual engagement with probabilistic unpredictability, black-box opacity, iterative refinement, and constitutive selection.

Publication Date

5-2026

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

COinS