Why Every AI Poster Looks The Same, And How To Fix It

You have seen these posters even if you never thought about them. Soft pastel colours. A slightly melty watercolour illustration. Bunting. A font trying very hard to be friendly. Cartoon children whose hands have the wrong number of fingers.

They are on every village fair notice board in Britain and every community bulletin board in America right now, and they all look like they came from the same place, because in a sense they did.

On Friday a blog post about this hit the Hacker News front page with 1,189 points and 661 comments, enormous attention for a post about flyers. The author, John Hartnup, was not writing an anti-AI rant. His argument was narrower: the tools can do far better, and everyone's poster looks identical because everyone asks the same lazy question.

If your kid sells bracelets, tutors, mows, bakes, or is about to pitch something at school, this is worth twenty minutes. The skill on display is not prompting. It is knowing what you want.

Because you asked for the average, and it gave you the average.

Ask an image tool for "a poster for a spring fair" with no other direction and it produces the most statistically typical spring fair poster it can imagine, assembled from every one it has seen. That is the machine doing its job. Hartnup is careful here: the problem is not that they are bad, it is that once you have seen that style twenty times it starts to irritate from sheer repetition.

Here is the part worth saying out loud to a fourteen-year-old. When everyone has the same tool and gives it the same instruction, nobody stands out. The tool has not made design easy. It made average design free, which means average design is now worth nothing, and the only thing with value left is the part the tool cannot supply: a point of view.

That is good news for a teenager with a business, because taste is cheap to acquire and expensive to fake.

Hartnup made up an event and started asking. The order is the method.

Attempt one. He gave the details and specified what he wanted: clean, unfussy, bright layout, bold spring graphic, and explicitly "avoid pastel/airbrush/oil style art or images of people." He got back something that still looked like what he was avoiding. Saying what you do not want is much weaker than saying what you do want.

Attempt two. A different instruction: make another using a completely different design aesthetic of your choice, and treat the current one as a "what not to do," because we want ours to stand out from posters that look similar. This time it produced something sharp and geometric, immediately better.

The key move. He asked what that style was called. It answered: Bauhaus-influenced modernist, geometric minimalism, some Swiss typographic style, with an explanation of each.

Now he had a word. Words are reusable. A style you can name is one you can ask for again, describe to a printer, or hand to a friend.

[IMAGE: Four event posters side by side in visibly different graphic styles, one geometric, one letterpress, one minimal, one collage] Alt text: The same event poster rendered in four distinct design styles Caption: Same information, four style instructions. The difference is the instruction, not the tool.

The most copyable trick in the whole post: instead of guessing at style names, ask the tool to list some.

Hartnup asked for a list of styles to choose from and got fifteen with descriptions. Some he had never heard of. Some he recognised but could not picture.

The list included risograph print, limited ink colours with slightly misaligned layers, handmade but designed. Cut paper collage in the style of Matisse, bold shapes that read from far away. Brutalist, raw and high-contrast and almost ugly on purpose. Japanese minimal, mostly empty space with one strong element. Wayfinding style, like transport signage. Memphis design, the squiggles and clashing colours from the 1980s.

This works for anything, not just posters. Ask for a menu before you ask for the thing. It turns a one-shot guess into a real choice, and choosing is where taste lives.

He later built a catalogue of a hundred poster styles with ready-to-paste prompts, a free education in design vocabulary that takes ten minutes to scroll.

Strip away the posters and a general rule sits underneath. The tool's default output is its average output. Its best output requires you to know something.

That applies to a poster, a logo, an essay outline, a business name, a product photo, a pitch script. The person who gets a great result is not the one with better software. Everyone has the same software. It is the person who can say "make it look like a 1960s Swiss transit poster" instead of "make it look good."

When you do not know the vocabulary yet, the move is not to give up. Ask the tool to teach you, then use it. Hartnup did not know what Memphis design was. He recognised it when he saw it, learned the name, and now owns that word permanently.

For a young person building anything, that loop is the whole game: notice you lack a word, get the word, use the word, keep the word.

One line near the end of Hartnup's post deserves more attention than it got.

You do not have to ask for a poster as a flat image. Tools like Claude and Gemini can produce HTML, PDF, or layered output where the text is real text. That means you can fix a typo, change a date, or swap a phone number without regenerating everything and hoping it comes back similar.

Dates change. Prices change. If your flyer only exists as a picture, every edit means starting over, and the new version never quite matches. Ask for the editable version. It is the difference between a poster and a template you can use all year.

One more note: the poster picked up extra text partway through because the tool invented a tagline for one style and carried it into every later version. New design, new chat.

If you have a teenager with something to advertise, here is the whole thing as a checklist. Have them do it, not you.

1. Write the information first, on paper. Name, what it is, date, time, place, price, contact. Poster design is mostly deciding what gets big and what gets small, and you cannot decide that until you know the words.
2. Ask the tool for ten poster styles suited to this kind of business, with a sentence on what each looks like.
3. Pick three that are nothing like each other. Not three variations of one idea.
4. Generate all three with the same information.
5. View them small. A poster's job is to work from across a room, and most designs that look great full screen disappear at thumbnail size.
6. Show them to three people who are not family and ask: what is this for? If they cannot answer in five seconds, the design failed however pretty it is.
7. Ask for the winner as editable HTML or PDF so the date can change later.

The same instinct applies to the rest of a small business: decide what the thing is before decorating it. If your teen is putting a real plan together rather than just a flyer, a school worksheet, a shared doc, or a structured platform like Foundra all work for getting the offer, customer and pricing written down. The format matters less than doing it before the marketing.

• AI poster tools default to the statistical average of every poster they have seen, which is why local event flyers everywhere now look identical.
• Saying what you do not want works poorly. Naming the style you do want works well.
• If you do not know style names, ask the tool for a menu with descriptions, then choose. Choosing is the skill.
• Ask what a style is called after you like something. A named style is reusable; a lucky result is not.
• Request editable output such as HTML or PDF with real text, so dates and prices can change without regenerating.
• Test any poster small and from a distance, and ask a stranger what it is for.

Is it wrong to use AI for a poster at all?

Hartnup's position is that you can be honest about using it; the goal is to avoid the look everyone is tired of. If you are advertising to a community that feels strongly about supporting local artists, hiring a person is a legitimate choice. For a teenager's first flyer with a budget of zero, using the tool thoughtfully is reasonable.

What age is this appropriate for?

Most image tools require users to be 13 or older, some 18 or parental permission. Check the terms of whatever tool you use, and consider having younger children work on a parent's account.

Will people be able to tell it was made with AI?

Often, yes. Hartnup says as much about his own results. The aim is not to hide it, it is to avoid producing the exact image twenty other events in your area produced this month.

What if my kid wants to learn design properly?

This is a decent doorway. Browsing a catalogue of a hundred named styles teaches more design vocabulary in ten minutes than most people pick up in a year. From there: library books on typography, and looking at posters in the wild and asking why they work.

What if the tool cannot get the text right on the image?

A known weakness, and exactly why the editable-file route matters. Generate the artwork, then place real text over it in a document or free design software. Never print anything without reading every word out loud first.

1. AI-generated posters don’t have to be horrible, John Hartnup, June 7, 2026
2. Poster Prompts: a catalogue of one hundred poster styles with example images
3. Hacker News discussion, 1,189 points and 661 comments, September 19, 2026
4. AI poster slop is taking over local events, The Independent
5. AI has taken over local event posters, and it shows, Meetings & Incentive Travel
6. Hacker News Digest, September 19, 2026