How to write prompts for AI to get better advertising designs?

Artificial intelligence can surprise you - one moment it delivers a design you would never have thought of, the next it returns something completely off the mark. The secret almost always lies not in the software, but in the prompt. The more precisely you describe what you want, the closer the result gets. In this guide we show, in plain language, how to build AI prompts so that instead of a random graphic you get an advertising design you can really use and send to print.

Why a specific description changes everything

AI does not read your mind - it works with what it is given. When you write "make a leaflet for my company", the model has to guess the industry, mood, colours and audience on its own. The result is often random, because it fills the gaps with the most common patterns. When you add that it is a leaflet for a dental clinic, in a calm, clean colour scheme, with one clear slogan and space for a phone number - suddenly you get something that makes sense.

The easiest way to think about prompting is like talking to a designer you are meeting for the first time. They do not know your company, your taste, and have not seen your logo. Everything obvious to you needs to be spelled out. The same goes for AI - the fewer things left unsaid, the fewer rounds of corrections.

What a good prompt is made of

You do not need to be a copywriter or know any jargon. It is enough for your prompt to answer a few simple questions. Treat it as a checklist you run through in your head before sending the prompt.

• What it is - a leaflet, an advertising poster, a business card, a post, a banner. The medium sets the proportions and the way the composition works.
• Who it is for - audience and industry. You speak differently to parents of preschoolers than to a premium salon client.
• What mood - elegant, cheerful, minimalist, energetic, calm. It is the shortest route to the right feel.
• Composition - what should stand out and where to leave room for the logo, slogan or contact details.
• Required elements - slogan, company name, address, icon, graphic motif. List them explicitly.
• What to avoid - specific things you do not want: "no photos of people", "no aggressive red", "no chaotic background".

The point is not to cram in every item. The point is to consciously pick the ones that decide the result in your case. Sometimes four sentences are enough, as long as they are specific ones.

A good prompt versus a weak one - with examples

It is easiest to see this side by side. Below is the same task in two versions. The first leaves AI too much freedom, the second guides it by the hand.

Weak prompt:

Make an advertising leaflet for a cafe. It should be nice and eye-catching.

It sounds reasonable, but it says almost nothing. What kind of cafe - a cosy local spot or a modern chain? What colours? What goes in the centre? The result will be a lottery.

Good prompt:

Design a portrait leaflet for a small cafe called "Green Corner". Cosy, warm mood, Scandinavian style. Colours: beige, warm green and wood. A large slogan at the top "Mornings taste better here", space for a photo of a cup in the centre, address and opening hours at the bottom. Plenty of light and space, no clutter. Avoid bright colours and stock photos of smiling people.

The second version gives AI a goal, audience, tone, palette, composition and a clear "no". Even if the first result is not perfect, you are much closer, and the fixes are cosmetic rather than fundamental.

Iteration - refine step by step

You rarely nail it on the first try - and that is perfectly normal. The strength of working with AI is that you can quickly ask for variants and polish the design in small moves. Instead of rewriting the whole prompt, add short corrective instructions.

• "Give me three variants of this layout, each with the slogan placed differently."
• "Make it warmer - more beige and wood, less cool white."
• "Fewer elements, keep only the slogan and one photo."
• "Change the layout - slogan at the bottom, photo at the top."
• "Enlarge the space for the logo in the top right corner."

Such a conversation is like refining a design with a graphic artist. Each round narrows the result and brings you closer to what you have in mind. It is worth changing one thing at a time - then it is easy to tell whether the tweak went in the right direction.

Describe the style in words and show inspirations

Style is usually the hardest part, because it is hard to name. And yet a lot depends on naming it. Instead of hoping AI hits your taste, describe it: "minimalist, lots of empty space", "1970s retro, warm colours", "elegant, thin lines, plenty of air", "hand-drawn, as if painted in watercolour".

Pointing to inspirations also helps - moods, eras, associations. "Feel like an old cinema poster", "like the cover of an elegant interior magazine", "cosy like a Scandinavian home". These are strong cues for AI that organise colours, typography and composition into a coherent whole. If you have your own logo or brand colours, state them explicitly and ask the design to match them.

Pitfalls to keep in mind

AI is a great helper, but it is not infallible - and it is worth knowing this in advance to avoid disappointment at the printing stage.

• Text on graphics can be wrong. Models often twist letters or add random characters. It is best to add wording yourself in a graphics program and let AI handle the background and composition.
• AI tends to make things up. It fills in elements from patterns, so it sometimes adds things nobody asked for. Always check the graphic for odd details.
• Image rights. Before using a graphic commercially, make sure what terms your tool sets. This matters when the material is heading to print and advertising.
• A human at the end. AI speeds up the start, but the final decision - whether the design is legible, coherent and print-ready - is made by a human.

If you want to calmly assess where AI really helps and where a human hand is needed, take a look at our post on whether artificial intelligence will replace the designer. It is a good starting point for setting realistic expectations.

Workflow: from prompt to a ready file

The healthiest approach is to treat AI as the first stage, not the whole job. A graphic from the model is a starting point you refine consciously. In practice it looks like this:

1. Prompt - you describe the goal, audience, tone, colours and format, and ask for a few variants.
2. Choice - you pick the direction that best fits your brand and message.
3. Refining in software - you fix the text, add the logo, tidy the layout and typography.
4. Preparing for print - you set the right format, bleeds, CMYK colours and resolution.

Most misunderstandings come from the last step. A nice image on screen is not yet a file ready for the press. How to prepare a graphic so it looks on paper the way it does in the preview, we explain step by step in our guide on preparing files for print. If you want to combine AI with a simple editor and a professional program in one process, we show it in the article on how to combine AI, Canva and a graphics program in one project.

Finally: the best prompt is your prompt

There is no single magic formula that always works. There is, however, a simple rule: tell AI exactly what you want to see, and do not be afraid to refine. The more specifics you give at the start and the more calmly you polish the result, the more often you get a design that truly fits your brand. And once the graphic is ready, we will make sure it looks just as good on paper as it does on screen.