top of page
Search

How to write better AI prompts (the thing nobody understands)

  • Jul 9
  • 6 min read

How to write better AI prompts for content creation

AI can't feel. We all know that. But most people don't realize how many of the words we use in our prompts are completely loaded with emotion, and that's exactly why your AI content comes out flat. Here's how to actually write prompts that work.

Everyone's writing prompts wrong, and it's costing them good content.

You sit down, you open ChatGPT or Claude, you write a good, detailed prompt, but somewhere along the way you type something like "write a useful post about skincare," and you get back something generic and lifeless. Then you blame the AI. The AI is dumb, the AI doesn't get it, the AI can't write like a human.


But the problem isn't the AI. The problem is your prompt.


I had a realization a while ago that completely changed how I use AI, and it's this: AI can't feel, and we all know that, but we don't actually realize how many emotions every single word carries.


Once that clicked for me, my prompts got dramatically better, and so did everything AI helped me create.


Let me explain what I mean, because this changes everything about how you should be writing your prompts.


Every word carries emotion


Think about the word "useful."


When you tell AI to "make this post useful," what are you actually asking for? You know what useful means. You've felt it. You've been in a situation where you needed something, you found exactly the right information at the right moment, and it solved your problem. That feeling of relief, of "oh thank god, this is exactly what I needed," that's what useful means to you.


Why AI cannot understand emotional words in prompts

AI has never felt that. It has never needed anything. It has never been stuck, frustrated, searching for an answer, and then found it. It has read a million examples of content labeled "useful" on the internet, so it can pattern-match what useful content usually looks like. But it doesn't know what useful actually is, because useful is a feeling, and AI can't feel.


So when you say "make this useful," AI does its best to imitate the pattern of useful content. But it's guessing. It's copying the shape of useful without understanding the substance, because the substance is emotional, and emotion is the one thing AI fundamentally cannot access.


This is the catch with using AI to make your content. The most important words in your prompts are usually the emotional ones, and those are exactly the words AI understands least.


The words that trip up AI


"Useful" is just one example. Think about how many words we throw into prompts that are secretly emotional:


Make it engaging. Make it punchy. Write something inspiring. Make it feel authentic. Write a compelling caption. Make it emotional. Make it relatable.


Every single one of these words is a feeling. Engaging is a feeling. Compelling is a feeling. Authentic is a feeling. You know what these words mean because you've experienced them. You've been engaged by something, compelled by something, moved by something authentic. AI hasn't.


So when you load your prompt with emotional words, you're asking AI to deliver something it can't understand. And it gives you back its best imitation, which is why it comes out feeling generic and hollow. It's copying the surface of emotion without the depth, because the depth requires having actually felt something.


This is why so much AI content sounds the same. Everyone's using the same emotional words in their prompts, and AI is giving everyone the same surface-level imitation of those emotions.


Take the emotion out AI prompt, then get specific


Here's what actually works.


You have to take the emotional layer out of your prompt, and replace it with concrete, pragmatic instructions. Instead of asking AI to make something useful, YOU decide what useful means, and you tell AI exactly what to write.


This is the key shift. You bring the feeling. You bring the judgment about what's actually useful, engaging, or compelling for your specific audience. AI brings the speed and the execution. You can't outsource the feeling to AI, but you can do the feeling part yourself and then hand AI clear, unemotional instructions.


Let me show you the difference.


Bad AI prompt vs good AI prompt example

Bad prompt:

"Write a post about skincare that will be useful to women in their 30s."


This is loaded with emotion ("useful") and vague. AI has to guess what useful means for women in their 30s. It'll give you something generic because you gave it a feeling to interpret instead of a task to execute.


Good prompt:

"Write a post about a skincare routine that includes X for the morning and X for the evening. Finish it with a CTA that calls women to download the full guide on our website. This post is meant for women in their 30s."


See what happened? You did the thinking. You decided what's actually useful for women in their 30s (a specific morning and evening routine). You took out the emotional word "useful" and replaced it with the concrete thing that makes it useful. Now AI doesn't have to guess what useful means. It just has to write what you told it to write.


You made the emotional decision. AI handled the execution. That's the whole game.


Why this works


When you take the emotion out and get specific, you're playing to AI's actual strengths.


AI is incredible at execution. Give it a clear, concrete task and it'll do it fast and well. Give it structure, details, and specific instructions, and it produces good work. What it can't do is make the emotional judgment calls about what will land with real humans, because it's never been a human feeling human things.


So you do the part AI can't do, which is decide what's useful, what's engaging, what's worth saying, who it's for and why they'll care. Then you give AI the part it's great at, which is turning your clear instructions into clean, fast output.


This is why the people getting the best results from AI aren't the ones with the cleverest prompts. They're the ones who understand their audience and their message so well that they can give AI extremely specific instructions. The thinking happens in their head. AI is just the typist.

How to actually write a good prompt


Beyond taking the emotion out, here's what makes a prompt actually work, based on everything I do when I use AI.


Give it context. Tell AI who you are, who the content is for, what your business does, what the goal is. The more context, the better it can match what you actually need. Don't make it guess the background.


Be specific about the topic. Don't say "write about marketing." Say "write about why businesses waste money running conversion ads before they have pixel data." The more specific the topic, the more focused and useful the output.


Tell it what NOT to do. This is huge. AI has default habits, like overusing em dashes, starting with generic hooks, structuring everything as bullet points, and sounding too formal. Tell it explicitly to avoid these. "No em dashes, no bullet points, write in flowing paragraphs, keep it casual."


Give it examples of your voice. If you want it to sound like you, show it how you sound. Paste in something you've written. AI is good at matching a style when you give it a reference, but it can't read your mind about how you want to sound.


Break big tasks into steps. Don't ask for a whole campaign in one prompt. Ask for the strategy first, react to it, then ask for the next piece. Working with AI in steps gives you way more control than trying to get everything at once.


Always edit the output. AI gives you a starting point, not a finished product. The emotion, the final polish, the voice, those come from you in the editing. Never publish AI output as it comes out.


The bigger lesson


This whole thing comes back to understanding what AI actually is and what it isn't.


AI is a tool that's brilliant at processing and producing language based on patterns. It's not a creative partner with taste, feelings, or judgment. When you treat it like the first thing, you get great results. When you treat it like the second thing, you get generic disappointment.


The businesses and people who'll thrive with AI are the ones who understand this distinction. They use AI for what it's good at, the execution, the speed, the heavy lifting, and they keep the human parts human. The strategy, the emotional judgment, the understanding of what actually moves people, the final voice. That stays with you.


If you want to dig deeper into using AI without losing your brand voice, I wrote a whole post about it here.


The point


AI can't feel. We all know this. But have you ever actually thought about how much emotion every single word carries?


When you fill your prompts with emotional words like useful, engaging, compelling, and authentic, you're asking AI to deliver feelings it has never experienced and cannot understand. That's why your content comes out flat.


The fix is to do the emotional work yourself. Decide what's useful, what's engaging, what matters. Then strip the emotion out of your prompt and give AI clear, specific, pragmatic instructions about what to write.


You bring the feeling. AI brings the speed.


That's the whole secret. 🤍


Want help using AI to work faster without losing what makes your business yours? We help small and medium businesses build smart, human content strategies. Schedule a free call and let's talk.

 
 
 

Comments


Subscribe to Our Newsletter

Thanks for submitting!

bottom of page