How to Write Better AI Prompts (No Tech Skills Needed)

I hear a version of the same complaint pretty regularly: "I tried ChatGPT and the answers were kind of useless."

When I ask what they typed, it's usually something like: "Write me a marketing email."

That's the whole prompt. And then they're surprised when the output is generic.

The tool isn't broken. The prompt is the problem. And fixing it doesn't require any technical skills. It just requires thinking a little more clearly about what you actually want.

Here's what I've learned after using these tools daily for the past couple of years.

The core problem: AI responds to what you give it

Think about hiring a new contractor. If you say "build me something in the backyard," they're going to ask you twelve clarifying questions before they do anything. A language model won't ask those questions. It'll just go build something and hand it back to you.

The output quality is almost entirely a function of how much context you provide upfront. This is the thing most people miss.

Give it a role before you give it a task

One of the fastest improvements you can make is telling the AI who it should be before asking it to do anything.

Instead of: "Write me a marketing email."

Try: "You're an experienced email marketer writing for a small law firm. Write a short email to clients announcing a new service we're offering."

Same task. Completely different results.

The role doesn't have to be elaborate. Even something like "you're a clear, no-nonsense writer" changes the output meaningfully. You're giving the model a frame to operate from.

Be specific about the output format

AI models love to give you more than you asked for. If you don't specify a format, you'll get whatever the model thinks is appropriate, which is often way longer and more structured than you need.

Tell it: - How long the response should be ("keep it under 200 words") - What format you want ("bullet points", "a short paragraph", "a numbered list") - What to leave out ("no headers", "skip the introduction", "don't include a call to action")

This sounds fussy, but it saves you from the editing work on the back end. It takes ten seconds to add to your prompt and cuts the revision cycle in half.

Include the context it doesn't have

The model has no idea who you are, what your business does, or who your customers are. You have to tell it.

If I'm asking Claude to help me write a follow-up email after a sales call, I give it: - What my company does - What the prospect said they needed - What I'm proposing to them - The tone I want (friendly but professional, not salesy)

Fifteen seconds of context. The difference between a generic follow-up and something I'd actually send.

A lot of people skip this step because it feels like extra work. But that extra work up front is the whole game. The model can only work with what you give it.

Use examples when the format matters

If you want the output to sound a certain way, show it an example of that style.

"Write this in a conversational tone" is vague. "Here's an example of the writing style I want: [paste a paragraph you like]" is concrete.

Examples work especially well for: - Subject lines and headlines - Social media posts where voice really matters - Any situation where you have a strong opinion about how something should sound

Paste two or three examples and tell it to match the style. You'll be surprised how well it calibrates.

When you don't like the output, ask it to try again

This is the part most people miss. One prompt, one output, then they give up.

AI tools are conversational. You can push back.

"That's too long, cut it in half." "The tone is too formal, make it friendlier." "I like the structure but the opening is weak, rewrite just the first paragraph."

You don't have to start over from scratch every time. Iteration is how you get from okay to good.

I probably revise three or four times on anything I'm going to publish or send. That's not the model failing. That's just how editing works.

A simple template to start with

If you want a quick structure you can use for almost any task:

1. Role: "You are a [type of expert or writer]."
2. Context: "[Relevant background about your business, audience, or situation]."
3. Task: "[Exactly what you want it to do]."
4. Format: "[Length, structure, what to include or exclude]."

You don't need all four every time. But when you're not getting what you want, run through the list and figure out what you left out.

The honest takeaway

Better prompts aren't a hack or a trick. They're just clear communication. The same clarity that makes a good email or a good meeting agenda makes a good prompt.

If you can describe what you want to a smart colleague, you can describe it to an AI model. The model will actually be more patient about it.

Start with one task you do regularly. Write a more specific prompt for it. See what happens. That's the whole experiment.