AI image prompts are written instructions that tell a tool like ChatGPT, Midjourney, or Gemini exactly what picture to generate. The best prompts follow a simple formula — subject + style + composition + lighting + details — and this page gives you 100+ copy-paste examples across portraits, products, logos, ads, real estate, food, and more.
A great AI image starts with a great prompt. Type “a dog” and you get something generic. Describe the subject, the style, the camera, and the light, and you get something you’d actually use on a website, an ad, or a product listing. Below is the formula that works in every tool, followed by 100+ ready-to-use prompts you can copy, paste, and tweak. If you’re new to prompting in general, start with our best ChatGPT prompts guide, then come back here for images.
The AI image prompt formula
Almost every strong image prompt is built from five parts. Add them in this order and your results jump instantly:
- Subject — what is in the picture (“a young Indian woman entrepreneur”).
- Style / medium — the look (“editorial photograph”, “3D render”, “flat vector illustration”, “watercolour”).
- Composition / shot — framing (“close-up”, “wide shot”, “top-down flat lay”, “rule of thirds”).
- Lighting / mood — (“soft natural window light”, “golden hour”, “dramatic studio lighting”).
- Technical details — (“85mm lens, shallow depth of field, 4K, hyper-detailed”, plus aspect ratio like
--ar 16:9in Midjourney).
Vague in, vague out. The more specific your words, the more usable your image.
Portrait & headshot prompts
Great for personal branding, LinkedIn, team pages, and author photos.
Product photography prompts
Perfect for e-commerce listings, ads, and catalogues.
Logo & brand prompts
Use these to generate starting concepts, then refine with a designer. (AI logos are a great brainstorming tool — always check for originality before you use one commercially.)
Social media & ad creative prompts
Thumb-stopping visuals for Instagram, Facebook, and ad campaigns.
Want these 100+ prompts as a ready-to-use pack?
Grab my organised AI prompt packs — categorised, tested, and formatted for ChatGPT, Midjourney & Gemini — from the shop. Or join the free list and I’ll send new prompt drops as I create them.
Browse the prompt packs →Real estate & interior prompts
For property listings, architecture firms, and interior designers.
Food & restaurant prompts
Make menus, delivery apps, and social posts look delicious.
Art, fantasy & illustration prompts
For book covers, games, blog art, and creative projects.
Which AI tool for which job
| Tool | Best for | Good to know |
|---|---|---|
| ChatGPT (image) | Quick, everyday images; understands plain English; great for beginners | Easiest to use; follows conversational prompts well |
| Midjourney | Artistic, high-aesthetic images; ads; concept art | Uses parameters like --ar, --style; steeper learning curve |
| Google Gemini | Photorealistic images; quick edits; integrated with Google tools | Strong with text-in-image and realism |
| Adobe Firefly | Commercial-safe images; design workflows | Trained on licensed content; good for client work |
To understand how AI tools are reshaping marketing as a whole, read AI in digital marketing and our guide to GEO (Generative Engine Optimisation).
Pro tips & mistakes to avoid
- Be specific, not long. Five precise details beat twenty vague adjectives.
- Name the medium. “Photograph”, “3D render”, or “flat vector” changes everything.
- Describe the light. Lighting is the single biggest quality lever (“soft window light”, “golden hour”).
- Set the aspect ratio for the platform (16:9 for banners, 9:16 for Stories, 1:1 for posts).
- Iterate. Generate, pick the best, then tweak one detail at a time.
- Check originality & rights before commercial use — especially for logos and anything resembling a real brand or person.
Tip: Save your winning prompts in a document with the tool and settings that produced them. A reusable prompt library is worth more than any single image.
- Use the formula: subject + style + composition + lighting + technical details.
- Copy the 100+ examples above and swap in your own subject.
- Match the tool to the job: ChatGPT for easy, Midjourney for artistic, Gemini/Firefly for realism & commercial work.
- Set the right aspect ratio for each platform.
- Always check originality and rights before commercial use.
Frequently asked questions
What are AI image prompts?
AI image prompts are written instructions that tell an AI tool like ChatGPT, Midjourney, or Gemini exactly what picture to generate. A good prompt describes the subject, style, composition, lighting, and technical details so the tool produces a usable, specific image instead of a generic one.
What is the best AI image prompt formula?
Subject + style/medium + composition + lighting/mood + technical details. For example: “Editorial photograph of a product on marble, soft side lighting, 85mm lens, shallow depth of field, 4K.” Adding these five parts in order dramatically improves results in any tool.
Which AI tool makes the best images?
It depends on the job. ChatGPT is easiest for beginners and everyday images, Midjourney is best for artistic and high-aesthetic work, and Google Gemini and Adobe Firefly are strong for photorealism and commercial-safe images. Many creators use more than one.
Can I use AI-generated images commercially?
Often yes, but it depends on the tool’s licence and your region’s laws, and you should always check for originality — especially for logos or anything resembling a real brand, product, or person. For client work, tools trained on licensed content (like Adobe Firefly) are a safer choice.
How do I write prompts for text inside an image?
State the exact text in quotes and keep it short, e.g. ‘a poster with the words “Grand Opening” in bold letters’. Gemini and newer models handle text-in-image best; still expect to regenerate a few times to get the spelling perfect.
Do longer prompts always give better images?
No. Clarity beats length. Five precise, well-chosen details usually produce a better image than a long list of vague adjectives. Start simple, then add one specific detail at a time until you get the result you want.