You’ve seen them. Taped to the pizza shop door, posted in the town Facebook group, tucked under your windshield wiper. The glossy flyer with the purple-and-teal glow, the weirdly perfect smiling family, the sparkles that came from nowhere, and a headline that says something like “Join Us for an Unforgettable Experience!”

Nobody reads past that. People clock it as AI slop in half a second and keep walking. Worse, it makes your business look like it didn’t care enough to try.

Here’s the thing: AI isn’t the problem. The prompt is. When you type “make a flyer for our fall festival,” you hand every creative decision to a machine built to give you the most average answer possible. And average is exactly what you get, the same average everyone else got.

This guide is about writing prompts that make AI work for you, so your flyer looks like it came from your business and not from the same button everybody else pushed.

Why every AI flyer looks the same

Image generators fill in every blank you leave with the safest, most popular choice. Don’t say what style, lighting, colors or layout you want, and the AI picks for you. It picks the same thing for everyone.

That’s where the tells come from:

• Glossy, plastic-looking 3D renders
• Purple, teal and orange gradients with glowing edges
• Sparkles, bokeh and lens flares for no reason
• Stock-looking people with too-perfect teeth (and sometimes too many fingers)
• Everything dead center, everything symmetrical, no room to breathe
• Gibberish text melted into signs in the background
• Copy stuffed with “unforgettable,” “elevate,” “vibrant” and “don’t miss out!”

Every one of those is a decision you didn’t make. Make them yourself and most of the slop goes away.

The six parts of a prompt that works

A good image prompt reads less like a wish and more like instructions you’d give a designer you just hired. Cover these six things:

1. Subject. What’s actually in the picture, with real details. Not “food trucks,” but “a taco truck with a hand-painted menu board and a line of people in hoodies.”
2. Medium or style. What it’s made of: screen print, linocut, watercolor, 35mm film photo, cut-paper collage, hand-painted sign. This one does the most work (more on it below).
3. Composition. Where things sit and where the empty space goes for your text. “Subject in the bottom third, plain sky across the top half for the headline.”
4. Color. Name a palette and keep it small. “Three colors only: brick red, cream and forest green.” Use your brand colors if you have them.
5. Mood and light. “Late afternoon, long shadows, a little gritty.” “Cold, foggy morning.” Light changes everything.
6. Limits. What to leave out, plus the shape you need: portrait for print, square or 4:5 for Instagram, tall for Stories.

You won’t need all six every time. But every one you skip is a decision the AI makes for you, and it’ll make the boring one.

Before and after: three real-world prompts

Brewery trivia night

Lazy prompt: “Make a flyer for trivia night at our brewery.”

What you get: Neon question marks, glowing beer mugs, a crowd of models clinking glasses.

Better prompt: “Two-color risograph-style print in mustard yellow and navy on off-white paper. A single pint glass on a worn wooden bar, drawn with slightly rough, offset ink like a cheap print job. Large empty area across the top third for a headline. Flat color, no gradients, no glow, no people, no text or lettering. Portrait, 8.5×11.”

Firehouse pancake breakfast

Lazy prompt: “Pancake breakfast fundraiser flyer, fun and family friendly.”

What you get: Cartoon pancakes with faces, rainbow sprinkles, a shiny 3D fire truck that looks like a toy.

Better prompt: “1950s diner-menu illustration style. A stack of pancakes with melting butter and a steel coffee mug on a red-and-white checkered tablecloth. Palette: fire-engine red, cream and black only. Bold, simple shapes like a hand-painted sign. Plate sits in the lower half; leave the top 40% plain cream for text. No text, no faces on food, no shine.”

Auto shop cars and coffee

Lazy prompt: “Car show flyer, cool and exciting.”

What you get: A chrome supercar, lightning bolts, flames and a sunset with lens flare.

Better prompt: “Black-and-white documentary photo shot on 35mm film with visible grain. Early morning in a strip-mall parking lot, a few older cars with hoods up, guys in work jackets holding paper coffee cups. Flat, overcast light. Shot low and wide, with clear sky across the top. No sports cars, no flames, no sunset, no lens flare.”

Notice none of the better prompts say “eye-catching,” “professional” or “stunning.” They describe a picture. That’s the whole trick.

Borrow a real style instead of asking for “eye-catching”

Words like “modern,” “eye-catching” and “professional” mean nothing to an image generator. They just push it harder toward the default look. Name an actual visual style with some history behind it instead. A few that work well for local flyers:

• Risograph print: grainy, two or three ink colors, slightly misaligned. Great for bars, shows, makers.
• Swiss poster style: bold grid, big simple shapes, lots of white space. Clean without being corporate.
• 1970s concert poster: warm oranges and browns, wavy shapes. Music, festivals, anything with a vibe.
• Hand-painted sign lettering: the old sign-painter look. Diners, shops, trades.
• Linocut or woodcut: chunky carved lines. Farms, craft beer, markets, anything handmade.
• Cut-paper collage: playful and textured. Kids’ events, schools, libraries.
• Mid-century travel poster: flat shapes, big skies. Shore towns, outdoor events.
• Photocopied punk flyer: black and white, high contrast, DIY. Shows, skate events, anything with an edge.
• Documentary film photography: real people, real places, grain and all. When you want it to feel honest.

Match the style to your crowd. A Saturday kids’ craft fair and a Friday-night metal show shouldn’t look like cousins.

One warning: skip “in the style of [famous living artist].” It looks derivative, it steps on a working artist’s toes, and everyone else is typing the same names. If you love a look, describe what you love about it: the colors, the texture, the way the type sits.

Tell it what not to do

The fastest way to kill the AI look is to ban it outright. Most tools listen to exclusions. Midjourney has a --no parameter, and others take a plain “do not include” line at the end of the prompt.

Keep a starter list and paste it every time:

No glossy 3D render, no gradients, no glowing edges, no sparkles, no lens flare, no bokeh, no text or lettering, no stock-photo smiles, no extra fingers.

Then add whatever tells show up in your first round. If every version comes back with a sunset, ban the sunset.

Keep the words out of the picture

Image generators still mangle small text, and text baked into an image can’t be fixed when the date changes or the price is wrong. So generate the artwork with no words and planned empty space, then add your headline, details and QR code in Canva, Adobe Express or whatever you use.

The exception: some tools, like Ideogram, handle a short headline pretty well. If you want stylized lettering as part of the art, keep it to one to four words and check it letter by letter. Dates, times, addresses and prices always get typed by a human.

Prompting the copy without the slop

The words have the same problem as the pictures. Ask ChatGPT or Claude to “write a flyer for our event” and you’ll get “Join us for an unforgettable day of fun for the whole family!” Every time.

Fix it the same way: give it the facts, a voice, a word limit and a list of banned words.

Write headline options for a flyer. Event: Cars & Coffee at Ray’s Auto, Saturday 8 to 10 a.m., free, Main St. lot. Audience: local car people, mostly 30 to 60, who roll their eyes at marketing talk. Voice: like Ray talking at the counter. Short, dry, a little funny. Give me 10 headlines under 7 words. Don’t use: unforgettable, elevate, vibrant, experience, don’t miss, join us, ultimate, exciting. No exclamation points. No emoji.

Then pick the closest one and rewrite it yourself so it sounds like you. Read it out loud. If you wouldn’t say it to a customer’s face, cut it.

A flyer only needs four things: a headline, what/when/where, one line on why it’s worth showing up, and what to do next. Anything else is clutter. And check every date, time and price against your own notes. AI will happily change them on you.

Iterate like a designer, not a slot machine

Hitting “regenerate” until something looks okay is how you end up with slop. Work it instead:

• Generate four or more versions, then change one thing at a time. “Same image, warmer light.” “Same image, move the glass to the left.” Small edits, not new rolls of the dice.
• Save the prompts that work. Build one house-style prompt (your medium, palette and exclusions) and reuse it for every flyer. Now your stuff looks like yours, not like everyone’s.
• Use reference images. Midjourney, Adobe Firefly and ChatGPT all let you upload an image to guide the style. Use your own photos, your logo colors, your storefront.
• Use real photos when you can. A real shot of your actual staff, shop or regulars beats a generated stranger every time. Let AI help with the layout and edits, not invent fake people.
• Do the across-the-room test. Print it, tape it to the wall, step back ten feet. If you can’t read the headline or the date, fix it before you print 500.

Steal this prompt template

Fill in the brackets and you’re ahead of 90% of the flyers on the corkboard:

[STYLE OR MEDIUM] of [SPECIFIC SUBJECT, with real details]. Setting: [where]. Light and mood: [time of day, feel]. Colors: [two to four named colors] only. Composition: [where the subject sits]; leave [area] empty for text. Shape: [portrait 8.5×11 / square / 4:5 / tall story]. Do not include: text, lettering, gradients, glow, sparkles, lens flare, [your own tells].

Before you print or post

• Could this flyer only belong to your business or event?
• Would someone glance at it and think “AI”? If so, find the tell and fix it.
• Date, time, address and price typed by you and double-checked
• Headline readable from across the room
• Text contrast checked with the WebAIM Contrast Checker
• QR code scanned and tested on a real phone
• No weird hands, warped signs or gibberish text hiding in the background

AI doesn’t make slop. Defaults do. Make the decisions yourself and let the machine do the grunt work.

Resources

• Midjourney documentation for parameters like --no and style references
• Adobe Firefly for image generation with reference-image matching
• Ideogram for art with short, stylized text
• WebAIM Contrast Checker to make sure people can actually read it