Prompting 101: Bad AI Output vs. a Great One

Most people think better prompting means finding the right words, some phrase or format that unlocks a smarter answer. That's not really it. A prompt isn't a spell. It's direction. And the gap between a bad prompt and a great one usually comes down to whether you actually gave the model the plan, vision, and voice we covered in the last post, or just asked it to guess.

Here's what that looks like side by side.

The bad prompt

"Write a blog post about the benefits of professional pool service."

Run that and you'll get something like this back:

"In today's fast-paced world, homeowners are busier than ever, and maintaining a clean, healthy pool can feel like a full-time job. That's where professional pool service comes in. From regular cleaning to expert chemical balancing, professional pool technicians offer a range of benefits that save you time and give you peace of mind."

Nothing in there is wrong. Nothing in there is yours, either. Swap the words "pool service" for "landscaping" or "HVAC maintenance" and the paragraph barely changes. That's the model doing exactly what it was asked: producing the average version of a topic it's seen described a thousand times, because that's all the prompt gave it to work with.

What's actually missing

Look at the bad prompt again. It doesn't say who's reading. It doesn't say what the business believes. It doesn't say what the piece needs to do, or what it should sound like. Four blanks, and the model fills every one of them with the safest, most generic answer available.

That's the same Vision, Plan, and Look and Sound breakdown from the last post, just applied one level down, to the prompt itself instead of the planning stage before it.

The good prompt

Same business, same topic, but written with direction instead of a guess:

"You're writing for a residential pool service company. Audience: homeowners who've had a green-pool scare, or just switched from another provider and don't fully trust chemical balancing to begin with. What we believe: most pool problems aren't bad luck, they're three preventable mistakes almost nobody explains to new pool owners. Write a post naming those three mistakes in plain language, no scare tactics, no chemistry jargon. Voice: direct, a little wry, like a technician talking to a neighbor over the fence, not a marketing department. Avoid phrases like 'in today's world' or 'when it comes to pool maintenance.'"

That prompt hands over an actual opinion, a specific reader, and a real voice instead of a topic. The draft that comes back still needs editing. It won't need reinventing.

A prompt template you can reuse

Steal this structure for your next piece:

"You're writing for [business], speaking to [specific audience or situation]. We believe [a real opinion about the problem, not a tagline]. Write about [topic] so it [what the piece needs to do: explain, reassure, persuade]. Voice: [two or three words describing tone]. Avoid: [the specific clichés or jargon your industry overuses]."

Fill in those blanks honestly and the first draft you get back will already sound closer to you than to a template.

A better prompt still needs an editor

One good prompt gets you a better draft. It doesn't get you a system. Someone still has to check the facts, cut whatever still sounds generic despite the direction, and make sure the piece matches the voice of everything else you've published. Doing that once is a skill. Doing it every week, across every channel, without it slowly drifting back toward average, is the part that takes more than a good prompt template.

Try the structure above on your next post. It'll get you further than you expect. Just don't mistake a better first draft for a finished one.

Previous
Previous

The SMB Owner's Guide to Getting Started with AI Content