If you’ve spent any time using ChatGPT, Claude, or Gemini, you’ve probably noticed something: sometimes the output is great, and sometimes it’s useless. Same tool, wildly different results. The difference almost always comes down to the prompt.
A prompt is simply the text you type into an AI tool. And just like a vague creative brief leads to a deliverable that misses the mark, a vague prompt leads to generic, unusable output. The good news is that writing better prompts isn’t complicated. It just takes a little structure and intention.
At Wormann Consulting, we work with AI tools daily across client campaigns, internal workflows, and strategic planning. These are the ten principles we come back to again and again.
How AI Actually Processes Your Prompt
Before jumping into tips, it helps to understand what’s happening on the other side of the screen. LLMs (Large Language Models) don’t think, reason, or retrieve facts from a database. They predict the most likely next word based on patterns learned from enormous amounts of text (If you’re interested in more info on how LLM’s work, see “What’s an LLM?“). Your prompt sets the context for that prediction. A specific, well-structured prompt narrows the range of likely outputs and pushes the model toward something useful. A vague one leaves the door wide open, and the model fills in the blanks with whatever generic pattern fits best.
That’s the entire game. Better input, better output. Here’s how to do it.
10 Tips for Writing Better AI Prompts
- Give It a Role
Start your prompt by telling the model who it is. “You are a senior public relations strategist preparing a media pitch for a mid-sized consumer brand launching a new project.” That one sentence changes the vocabulary, depth, and assumptions the model uses. Without a role, you get a generalist. With one, you get a specialist.
- State the Output Format Explicitly
LLMs default to long prose. If you want bullet points, a table, or a short paragraph, say so. “Return five bullet points, each under 25 words, no intro sentence.” Think of it like a creative brief. You wouldn’t hand a designer a vague description and expect exactly what you had in your head.
- Front-Load Your Most Important Constraint
Here’s something most people don’t realize: words earlier in the prompt carry more influence. If the audience, tone, or word count matters, put it in the first sentence rather than burying it at the end.
- Show It What Good Looks Like
One concrete example outperforms three paragraphs of abstract description. Paste in a sample of the format, tone, or structure you want and tell the model to match it. Upload a similar previous file, give it a web page, or copy/paste a volume of text that feels similar to what you want produced. This is called few-shot prompting and it’s one of the highest-leverage techniques available.
- Tell It What NOT to Do
Negative constraints are underused. “Don’t use jargon. Don’t assume a large budget. Don’t recommend tools that require a developer.” These cut off common failure modes before they happen and keep the output grounded in your actual situation. Of course, don’t forget one of the most important prompts: “Don’t use em dashes.”
- Ask It to Think Step by Step
For complex or analytical tasks, add: “Think through this step by step before giving your final answer.” This forces the model to build a reasoning chain instead of jumping to the first plausible response. For simple tasks it’s overkill, but for strategy and analysis, the difference in quality is noticeable.
- Use Labeled Sections to Organize Complex Prompts
When a prompt includes instructions, context, and data all at once, the model can start blending them together. Wrapping each section in labeled tags (like <instruction>, <context>, and <data>) draws clear boundaries. The model treats each section according to its label, which keeps your instructions from bleeding into your content.
- Include the “So That” Clause
“Summarize this report so that a non-technical client can understand the key risk without reading the full document.” That “so that” changes word choice, depth, and framing in ways that format instructions alone can’t replicate. Purpose shapes output. Tell the model what success looks like for the person on the receiving end.
- Iterate Instead of Starting Over
A mediocre first output is usually one follow-up away from something usable. “The tone is right but it’s too long. Cut it by half and lead with the risk.” That’s not failure. That’s the process. And when you land on a prompt that works reliably for a recurring task, save it. Over time, a library of proven prompts becomes a real operational asset.
- Understand Temperature
Temperature controls how “creative” the model’s word choices are. Low temperature produces consistent, precise, predictable output. High temperature produces more varied, surprising results. In API environments, it’s a literal number you set. In chat tools, you can approximate it with language: “Give me the single most accurate answer” nudges precision, while “Give me five wildly different angles” nudges creativity.
And finally, don’t forget to double check the output you get. LLM’s are great tools, but they’re not perfect and they have been known to hallucinate some information. Always fact check the outputs you get, and don’t be afraid to challenge the output. Training has biased LLM’s toward what you want to hear, so you may get incorrect information that feels good as opposed to the cold hard truth. Pushing back can often get you the real answers, which are what you need.
The Bottom Line
AI tools are becoming standard operating equipment for modern businesses. But the tool itself is the same for everyone. The difference is how you use it. A well-crafted prompt turns a generic chatbot into a strategic asset. A careless one turns it into a liability.
The organizations that get the most out of AI won’t be the ones chasing every new feature announcement. They’ll be the ones that build practical, repeatable processes around how they use these tools every day.
Have some questions? Need a partner that can help you with your marketing and beyond? Reach out to start a conversation.
