ChatGPT Images 2.0 inserted a road sign reading “WHY ARE YOU LIKE THIS” into a generated image without being asked to do so, according to a post by developer Simon Willison.
The prompt that produced the image, sent by @scottjla on X, was: “Create an image of a horse riding an astronaut, where the astronaut is riding a pelican that is riding a bicycle. It looks very chaotic but they all just manage to balance on top of each other.”
The resulting image depicted the described scene — a pelican on a bicycle, an astronaut on the pelican, a horse on the astronaut — with a road sign in the background bearing the phrase, which did not appear in the prompt.
Willison said he verified that the model added the text of its own accord. He linked to a shared ChatGPT conversation at chatgpt.com confirming the same prompt produced the sign.
The post is part of Willison’s ongoing “pelican riding a bicycle” benchmark series, which he uses to informally track and compare image generation model outputs across time. The incident is a single observed output, not a systematic evaluation, and Willison did not claim it indicates a broader pattern.
This article is based on a single tier-1 source.