preprint 2026 ยท arXiv

Prompt and Circumstances: Evaluating the Efficacy of Human Prompt Inference in AI-Generated Art

K. Trinh, S. Seidenberger, J. Spracklen, R. Wijewickrama, B. Viswanath, A. Maiti

arXiv

Abstract

Evaluates the efficacy of human prompt inference in AI-generated art. Through controlled experiments, we assess whether humans can reliably determine the prompts used to create AI-generated images.

Key Results

  • Human prompt inference accuracy is low and highly variable
  • Contextual information significantly affects inference performance
  • Results have implications for AI art attribution and intellectual property