How to Write AI Image Prompts: The Six-Part Method
How to write AI image prompts for your business: skip the magic words and cover six parts, from Subject to Brand Cues. Examples included.
Search "best AI image prompt" and you'll find lists of magic words. "8K. Ultra-detailed. Award-winning. Masterpiece. Cinematic." People paste them in and hope.
Those words might add a little shine. They tell the tool nothing about your business. You get a shiny, generic picture that could belong to anyone, and then you decide AI images don't work.
If you want to know how to write AI image prompts that actually work for a business, forget the magic words. The question isn't what to say. It's what to cover. I use a structure I call the Six-Part Image Prompt, and it's built on one idea: answer the questions the tool would ask you, if it could.
Coverage beats length, and magic words beat nothing
Think about walking into a bakery and saying, "I want a cake." A good baker has a hundred questions. What kind? How big? What's the occasion? What colors? Should it say anything?
An AI image tool doesn't ask. It guesses. And every guess is the most common answer it knows.
So when you skip a part, the tool fills it in for you. Skip the style and you might get a cartoon when you wanted a photo. Skip the framing and your product lands dead center with no room for a headline. Skip your brand details and you get someone else's colors.
The second myth is that longer prompts are better. They aren't. A long prompt full of fuzzy words still leaves the tool guessing. A shorter prompt that covers all six parts usually wins.
The six parts of an image prompt
Each part answers one question:
| Part | The question it answers |
|---|---|
| Subject | What is the main thing in the picture? |
| Scene | Where is it, and what's around it? |
| Style | What kind of image? Photo, illustration, flat graphic? |
| Composition | How is it framed? How close? Where's the empty space? |
| Lighting & Color | What's the light like, and which colors lead? |
| Brand Cues | What makes it look like you, and what must never appear? |
Here's what each one needs.
Subject. Name the thing exactly, say how many, and say what it's doing. "A dog" is weak. "A freshly groomed golden doodle sitting and looking at the camera, with a small blue bow on one ear" is strong. Pick one main subject. Two equal subjects fight for attention.
Scene. Name the place and add one or two background details that fit your business. "In a kitchen" is weak. "On a worn wooden bakery counter with a blurred chalkboard menu behind it" tells the viewer what kind of place you are. Keep the background simple if the subject needs to stand out.
Style. Most small businesses do best with one of three: photo-style (feels real), illustration (feels friendly), or flat graphic (clean, works well behind text). Pick one per image. Mixing "photo" and "cartoon" in one prompt gets you a strange blend.
Composition. This is the part beginners skip most, and it's the part that decides whether the image fits where it's going. Say the distance (close-up, medium, wide), the angle (eye level, from above, straight down), where the subject sits (left third, right third, center), where you need empty space for words, and the shape (square, tall, or wide). If your tool has a size setting, use it too.
Lighting & Color. Light sets the mood faster than anything. Warm, low light feels cozy. Bright, even light feels clean and professional. Then add your brand colors by name and hex code, like "navy (#1B2A4A)." Some tools follow names better than codes, so give both.
Brand Cues. "On brand" means nothing to a tool. It doesn't know your brand. Spell it out: two or three things that always show up for you ("handmade feel, real flour dust, imperfect shapes") and your biggest don'ts ("no text, no logos, no faces"). If your tool has a separate box for things to avoid, put the don'ts there.
What a six-part prompt looks like in practice
Say a dog groomer wants an Instagram post announcing nail trims. Her brand is bright, playful, sunny yellow and sky blue. Her first try was:
a dog at a grooming salon
She got a serious poodle on a steel table in a gray room. It felt like a vet's office. Fine picture. Wrong business.
Here's the same idea, built part by part:
| Part | What she wrote |
|---|---|
| Subject | A happy golden retriever holding up one front paw like a high-five while a smiling groomer gently trims its nails |
| Scene | A bright, cheerful grooming salon |
| Style | Realistic photo, cheerful and candid |
| Composition | Medium shot at the dog's eye level, square, open space at the top for a headline |
| Lighting & Color | Soft, even daylight; sunny yellow and sky blue walls with white trim |
| Brand Cues | Playful, friendly mood; no text in the image; no logos |
Stitched together, it reads as a few plain sentences. Every phrase answers a question the tool would otherwise guess at. What dog? Doing what? In what room? What colors? What mood? What shape?
The real payoff comes later. When an image works, you know why. Next month you swap the Subject, keep the other five parts, and the new image looks like it belongs next to the old one. That's how a feed starts to look like one business on purpose instead of by luck.
Don't ask the generator to write your words
One rule saves a lot of frustration. Notice that her Brand Cues say "no text in the image." That's on purpose.
Image generators still get text wrong. Letters bend, words come out misspelled, fonts look odd. If you need "Nail Trims Now Available" on the post, ask the generator for empty space where the words will go, then add the words yourself in a design tool where you can type and check them. Picture in the generator. Words in the design tool. Every time.
The honest caveat
The Six-Part Image Prompt doesn't make the first result perfect. AI tools make something different every time, even from the same prompt. You'll still look at a few drafts, pick the closest one, and adjust. The six parts make that easier, because when something's wrong you know which part to change. Wrong mood is usually Lighting & Color. Subject too small is Composition. Looks generic is usually Scene or Brand Cues.
Tools also differ. Some follow color codes well, some ignore them. Some handle long prompts, some get lost. The six parts still apply. You just learn what your particular tool listens to.
And a prompt can't fix a missing decision. If you don't know who the image is for or what it's supposed to do, no structure will save it. The six parts work best when you've already decided the job and written down your brand's look. Then the prompt mostly writes itself.
Key takeaways
- Magic words add shine but tell the tool nothing about your business.
- A Six-Part Image Prompt covers Subject, Scene, Style, Composition, Lighting & Color, and Brand Cues.
- Coverage beats length: a short prompt with all six parts usually beats a long fuzzy one.
- Composition is the part beginners skip, and it decides whether the image fits its channel.
- Make the picture in a generator and add the words in a design tool.
Frequently asked questions
How do you write a good AI image prompt for a business?
A good business image prompt covers six parts: the Subject (what's in the picture), the Scene (where it is), the Style (photo, illustration, or flat graphic), the Composition (framing, shape, and empty space for text), the Lighting & Color (light and brand colors), and Brand Cues (what makes it yours and what must never appear). Covering all six leaves the tool far less to guess.
Do magic words like 8K or masterpiece improve AI images?
Words like 8K, ultra-detailed, or masterpiece may add some polish, but they don't tell an AI image tool anything about your business, customers, or brand. The result is a shiny but generic image. Specific details about the subject, setting, framing, lighting, colors, and brand rules do far more to make an image useful.
Why does AI misspell words in images?
AI image generators build pictures by predicting what pixels should look like, not by typing letters, so text inside an image often comes out bent, garbled, or misspelled. The reliable fix is to ask the generator for empty space where words will go, then add the text yourself in a design tool such as Canva or Microsoft Designer, where you can type and proofread it.
How long should an AI image prompt be?
Long enough to cover the subject, scene, style, composition, lighting and color, and brand cues, and no longer. Length alone doesn't help; a long prompt full of vague words still leaves the tool guessing. For most small business images, a few plain sentences that cover all six parts work better than a paragraph of adjectives.
Where this goes next
The Six-Part Image Prompt is the heart of Module 2 in How to Use AI to Create Images for Your Business, where you take it from prompt to a finished, checked, ready-to-post image. If you want the whole path, that's what the class covers.
See the class