Instead of using Google as usual, they tried ChatGPT, typing simply: “Things to do in Hot Springs.”

The results? Technically correct, but generic – like a stack of TripAdvisor brochures left out in the rain. The AI had no idea about their preferences: quiet trails over tourist traps, locally roasted coffee, restaurants with unapologetic pie.

The realization was profound: the model didn’t fail; they just hadn’t told their story.

Why AI Feels Like It Misses the Mark

If you’ve ever gotten a response from ChatGPT that felt weirdly specific yet completely unhelpful, you’re experiencing what’s called by some “the mirror problem.” These tools aren’t mind readers – they’re mirrors reflecting how well we express ourselves.

The solution comes from an unexpected place: journalism. Those old newsroom questions – who, what, why, and how – turn out to be perfect for crafting AI prompts that actually deliver.

The Four-Part Framework for Better Prompts

1. Who You Are

Every effective prompt starts with context. You don’t need your life story, but you do need to establish perspective.

Are you a teacher designing an engaging lesson? A project manager prepping for executives? A parent settling a debate about penguins? The more you reveal about your role, goals, or situation, the sharper the response becomes.

Why this matters: AI models predict language patterns. Without a starting point, they’re guessing at your intent. Think of it as handing the model a pair of glasses – without them, everything stays blurry.

2. What You Want Done

This is where most people lose clarity. We assume our intent is obvious. “Write something about marketing trends” or “Give me trip ideas” feel specific enough, but they’re not.

“Something” and “ideas” are slippery words.

Lead with a strong verb instead. Do you want the AI to analyze, compare, brainstorm, design, or narrate? Large language models are like Swiss Army knives – capable but aimless until you select the right tool.

Compare these requests:

• Vague: “Give me vacation ideas”
• Clear: “Create a three-and-a-half-day itinerary for a couple visiting Hot Springs who love nature and local food”

The second gives the AI a mission, not a riddle.

3. Why You Want It

This is the step most people skip, yet it makes the biggest difference.

When you explain your motivation, you teach the AI how to weigh trade-offs. Are you trying to impress someone? Learn something new? Persuade stakeholders? Save time?

Search engines never cared about your “why,” but language models can adapt tone, detail, and format based on intent.

Clarity of purpose drives quality of response. Asking for vacation ideas to surprise a spouse will emphasize romance and originality. Researching for a client shifts the focus to evidence and structure. Sharing your motive helps the AI find the heartbeat behind your request.

4. How You Want It Styled

Once you’ve set the stage, you get to direct the performance. This is where AI becomes a true collaborator.

Do you want conversational or executive tone? Poetic or procedural? Should it sound like a local guide, sports announcer, or enthusiastic friend?

Style makes information stick. The same facts can read like a corporate memo or a campfire story depending on how they’re presented. When you specify tone, you’re not micromanaging – you’re setting the mood that makes your content resonate.

Putting It Into Practice

Here’s what all four elements look like together:

“I’m a middle-aged husband planning a four-day anniversary trip to Hot Springs, Arkansas, this spring. Lodging is already handled. I’d love a three-and-a-half-day itinerary of hikes, scenic views, local shops, and mid-range restaurants – something that mixes nature and indulgence. Nothing too fancy, but we’re not eating gas-station tacos either.”

See the difference? It’s not a search query. It’s a story with a goal.

The Final Touch: Leave the Door Open

The best prompts don’t end with a period – they end with curiosity.

When you’re uncertain about clarity, add a simple question: “Does that make sense?” or “What else would you need to know before you start?” This isn’t showing weakness; it’s modeling collaboration. The AI pauses, considers, and often fills in gaps you didn’t realize existed.

You can also add stylistic flair at the end:

• “Make it sound like a late-night monologue”
• “Write this as if we’re old friends catching up”
• “Give me three ideas that would make my boss grin”

These tags act like punctuation marks, cueing tone and direction while reminding the model you’re co-creating, not dictating.

The Takeaway

When you approach AI with the who, what, why, and how framework, you transform these tools from disappointing search alternatives into genuine collaborators. You’re not extracting information – you’re co-creating solutions.

Remember: AI systems are mirrors. They reflect the clarity of your communication. Give them your story, your purpose, and your voice, and watch the quality of responses transform.

The next time you sit down to write a prompt, don’t just ask a question. Tell a story with a goal.

---

Tip from me – if you don’t know how to craft the “who, what, why, and how”, use the AI to generate the prompt for you, eg :

• You : “Things to do in Hot Springs, please give me one example prompt, with suggested context – who, what, why, and how”
• AI : “Create a detailed weekend itinerary for a small group of friends in their early 30s who are history buffs and wellness enthusiasts visiting Hot Springs, Arkansas. Include a mix of outdoor activities at Hot Springs National Park, a guided tour of the historic bathhouses, local dining recommendations featuring Southern cuisine, and tips for soaking in thermal waters. Explain why each activity appeals to this group’s interests, how to get around town, and suggestions for budget-friendly accommodations.”

(Actual prompt and reply using Perplexity).

You can iterate this process to refine the prompt, AI assisted, down to exactly what you want to ask, with “context”, just keep on refining it until it’s “perfect”.

Using this approach you can customise your prompt, to whatever you want to get quality information for. Basically, leverage the AI, to help you write the prompt, then send the “context rich” prompt to the AI.

Inspired by an article from Chris Murman @ substack, thanks Chris.