I'd accept that, and laud it as a valid use case for AI images even; as long as it was clear what was being done and had a human in the loop to ensure the generated image was reasonable.