Five posters made the top five at the Ohio State Fair’s poster contest. The one that won was made with AI, and it carried the usual tells: the common image errors anyone who has spent ten minutes with an image generator can spot from across a room. Judges either missed them or didn’t care. Ohioans noticed, and they were not thrilled.
That single contest result tells you most of what you need to know about where we are with generated design work. The tools are capable enough to win a competition. They are not capable enough to survive a close look. And somewhere in that gap sits every ugly flyer currently stapled to a coffee shop bulletin board.
We are, apparently, in a flyer pandemic
The phrase going around is “ChatGPT flyer pandemic,” and it is accurate in the way most internet phrases are not. These posters have spread into social media feeds, bulletin boards, restaurant windows. When one outlet asked readers to send in the worst examples they had encountered, the submissions came in. Plenty of them.
What makes the pile-on interesting is the shape of the complaint. The loudest objection is not philosophical. It is aesthetic. As one write-up put it, the AI versus not-AI argument is one thing, but regardless of how these posters were made, they are all just ugly. That is a much harder criticism to wave away than any argument about training data or artistic labor, because ugly is measurable by anyone with eyes and no opinion about machine learning.
I review AI tools for a living. I am not in the business of pretending these systems are useless. But I am also not going to sit here and tell you the average generated poster is fine. It is not fine. The type is mushy. The hierarchy is nonexistent. The lighting comes from four directions at once. There is often a hand doing something a hand cannot do. These are not edge cases. They are the default output of people who asked a model for a poster and accepted the first thing it handed back.
The fix is not a better model
The most encouraging development in this whole mess is that human designers have started taking these disasters and rebuilding them. Not mocking them, not writing think pieces about them, actually fixing them. It is framed as AI versus humans, and in that framing the humans are winning.
I want to be careful about what that proves, because it is easy to misread as an argument that the tools are worthless. It is not. What it proves is that the bottleneck was never generation. Generation is solved enough. The bottleneck is judgment: knowing that a poster needs one focal point, that a headline has to read at six feet, that the eye needs somewhere to rest. None of that is in a prompt. All of it is in a trained designer’s head.
Which is why the “AI ruined posters” headline is slightly off, and I say this as someone predisposed to be hard on these tools. AI did not ruin posters. Cheap indifference ruined posters, and AI made cheap indifference faster to execute. A person who did not care about design in 2019 made a bad poster in Word. That same person now makes a bad poster in thirty seconds and shares it wider. The volume changed. The taste did not.
What this means if you actually use these tools
The practical takeaway for anyone shipping visual work with AI in the loop:
- Treat the first output as a rough draft, not a deliverable. If you ship what the model gave you unedited, you are shipping the thing people are now collecting as examples of bad work.
- Fix the type yourself. Generated text and generated typography are where these images fail most visibly and most often, and they are also the easiest part to correct in any real design tool.
- Look at the hands, the reflections, the background text, and the edges. The Ohio State Fair winner had common errors in it. Someone checking for thirty seconds would have caught them.
- If the work carries your name or your organization’s, have a person with design training touch it before it goes out. Not as a formality. As quality control.
The uncomfortable part for the contest judges
The Ohio story is not really an AI story. It is a judging story. A poster with visible generation artifacts beat four other entries, and the resulting controversy was not about whether AI should be allowed. It was about how something with obvious flaws came out on top. That is a question for the people holding the clipboards.
Human creativity is still the thing being valued here, and the market is making that clear in the least flattering way possible: by circulating the worst AI output as entertainment while paying designers to clean it up. Generated posters do not have to be horrible. They are horrible when nobody with taste is in the room.
🕒 Published: