I’ll admit something. For the longest time I treated AI prompts like a legal contract. A giant wall of “don’t do this” and “never say that.” And the output? Still sounded like a robot wearing a suit. So when I came across this post from an AI professional breaking down a totally different approach, I stopped scrolling.
The author nailed a problem I see everywhere. Most people prompt AI like they’re writing a terms and conditions doc. A list of things to avoid. No examples. No voice. No reference point. Just 30 rules telling the model what NOT to do, and then everyone wonders why every output sounds identical.
I love how plainly the original poster put it. Rules tell the model what to avoid. They don’t show it what to do. That one distinction is the whole ballgame.
Why “don’t” lists keep failing you
The creator watched this play out on their own team over and over. People would spend ten minutes loading a prompt with restrictions. “Don’t use buzzwords.” “Don’t start with ‘In today’s world.'” “Don’t sound like AI.” The output would still sound like AI.
Here’s the part that clicked for me. A model can follow every restriction you give it and still produce something flat, because you’ve only told it what bucket to stay out of. You never handed it a target to aim at. Restrictions remove options. They don’t create a voice.
The fix: feed it a reference file
Instead of rules, this savvy professional flips the whole thing around. You give the model a reference file. Ten to fifteen pieces of your actual writing. Real posts. Real emails. Real threads. The ones you’re genuinely proud of.
The model reads them and picks up the stuff a rule could never capture: your sentence rhythm, your vocabulary, how you open, how you close. Then you write one line in the prompt: “Apply everything you learn as a writing rulebook.” That’s it.
The exact step-by-step build
This is the part I think you’ll want to save. The expert laid out the full process for building a reference file that actually works, and each step has a reason behind it.
- Gather 10 to 15 pieces of writing you’re proud of. Posts, emails, threads. This is the raw material the model learns from, so quality beats quantity.
- Paste them all into one document. One file means one upload and one clean source of truth, not a dozen scattered samples.
- Label each piece with a one-line note on why it worked. Things like “High engagement, conversational tone” or “Short sentences, strong opening.” This tells the model what to pay attention to, not just what to copy.
- Add a short section at the top describing your voice in plain language. Five or six sentences max. It gives the model a quick anchor before it even reads the samples.
- Upload this file every time you start a new writing session with AI. Consistency is what keeps every draft sounding like you instead of drifting back to generic.
- Tell the model to study the file before writing anything, and ask it to ask clarifying questions first. That pause forces it to absorb your voice instead of rushing to a bland first draft.
Why it matters: rules constrain the model into a corner. A reference file gives it a living example of how you actually sound. One subtracts, the other teaches.
The before and after
The shift the original poster described is dramatic, and it lines up with what makes this so practical.
- Before: ten minutes writing restrictions into every single prompt, and still getting generic output.
- After: thirty seconds uploading one file, and the first draft already sounds like you.
I think this is so useful because it scales. Build the reference file once, reuse it forever. And it works whether you’re a solo creator or running a whole content team. Anyone still stuck on rules-only prompts is basically doing ten times the work for a worse result.
It also connects to a bigger trend I keep noticing. The people getting real value out of AI right now aren’t the ones with the cleverest restrictions. They’re the ones feeding the model rich context and real examples. Show, don’t tell, turns out to be just as true for machines as it is for writers.
One small tip from me as you try this: keep refreshing your reference file as your writing evolves. Swap in newer posts you’re proud of, and the model’s sense of your voice stays current instead of frozen in last year’s style.
The mind behind this post also packed a full visual breakdown into an infographic, so if the step-by-step clicked for you, the original LinkedIn post is worth the read. Check it out for the complete walkthrough, and maybe pass it along to someone on your team who’s still running on rules-only prompts.