Those hideous AI flyers clogging your neighborhood bulletin board aren’t proof that AI can’t design. They’re proof that the person who made them didn’t know what a poster is supposed to do. The tool is not the problem. The absence of taste is the problem.
I say this as someone who spends most of his working hours testing these tools and writing down what breaks. I have generated enough bad output to fill a landfill. And the failure pattern is boringly consistent: the model does exactly what it is told, the person telling it has no idea what to ask for, and the result gets posted anyway because it took eleven seconds and cost nothing.
What the flyer pandemic actually shows
There is a genuine wave of this stuff. Publications have been collecting reader-submitted examples of AI-generated posters and advertisements that have spread across social media, bulletin boards, and restaurant windows. One writer described it as a “ChatGPT flyer pandemic.” The most damning critique in that coverage wasn’t about ethics or authorship. It was simpler than that: regardless of how these posters were made, they’re all just ugly.
That’s the review I trust most, because it skips the ideology and goes straight to the output. Ugly is a measurable failure. And ugly is not a property of the model. It’s a property of the workflow.
Think about what most of these flyers are. Someone needs to advertise a car wash or a church potluck. They type a description into a chat box, take the first image that appears, slap unreadable text on top of it, and print thirty copies. No second pass. No editing. No layout. No sense of hierarchy, contrast, or whether the phone number is legible from four feet away. The same person with Microsoft Word and a clip-art CD in 2003 would have produced something equally bad. The tool got faster. The judgment didn’t arrive.
The contest problem is a judging problem
The Ohio State Fair situation makes the case better than I can. A poster created with artificial intelligence won the fair’s poster contest, and Ohioans raised concerns. What made it worse was that the winning entry contained common AI-generated image errors. Not subtle ones. The recognizable tells.
So a piece with visible defects beat human-made entries. That’s not a story about machines out-designing people. That’s a story about a judging panel that either couldn’t spot the errors or didn’t care to look. If the winner had flaws that the public could see from a news photo, the evaluation process failed before the technology did. Blaming the generator for that is like blaming a calculator for a bad accountant.
Humans fixing the output is the actual lesson
There’s a designer currently making a project out of repairing these posters, and the framing around that work is “AI versus humans, and the humans are winning.” I like the work. I don’t love the framing, because it describes a collaboration as a fight.
What that designer is doing is the missing step. They’re supplying the thing the original maker skipped: a trained eye deciding what stays, what goes, and how the whole thing reads at a glance. That’s not a victory over the tool. That’s a demonstration of what the tool needs to be useful.
Every AI toolkit I’ve reviewed that produces decent visual work has the same shape:
- A person who knows what good looks like before generating anything
- Generation used for raw material, not finished deliverables
- A real editing pass in real design software, where type and spacing get handled by a human
- Willingness to throw out the first twenty results
The toolkits that produce garbage skip every one of those steps. Same models. Wildly different outcomes. That gap is the whole review, honestly. When I test a generator and the output is unusable, I’ve learned to check my own prompt and my own standards before I write the tool off.
What I’d tell you to do
If you’re making a poster and you have no design training, AI will not substitute for the training. It will accelerate whatever instinct you already have, including the bad instincts. Use it to explore directions quickly, then hand the result to someone who can finish it, or learn enough typography to finish it yourself. Text on images is still where these tools embarrass you fastest, so set the type manually.
If you’re judging a contest, look closely at the entries. The errors are findable. Ohio proved they’re findable by anyone with a news article and thirty seconds.
The current consensus is that human creativity is valued over AI in design, and I agree with it, but not for the reason most people cite. It’s not that humans make prettier pixels. It’s that humans are the only part of the pipeline capable of deciding whether something is any good. Remove that and you get a bulletin board full of nonsense. Keep it and these tools stop being a punchline.
🕒 Published: