Teaching AI to Sound More Like You is a Process...
... and goes WAY beyond the prompt
The TL;DR (Executive Summary)
AI will not magically “know your voice” because you asked it to. You have to teach it.
The real learning happens when you ask AI to compare its draft to your edited, complete version.
Your edits reveal more than word choice. They reveal your rhythm, priorities, humor, structure, and point of view.
The best AI writing process is not “prompt and publish.” It’s draft, revise, compare, teach, and repeat.
The true “magic” comes in when you ask the AI what it learned after comparing its draft to your final edited copy.
Human in, better human out: Still undefeated.
I recently had one of those “off-the-cuff” moments that turned out way better than I ever expected. First, some background.
One of the most common things people want from AI is for it to write in their voice. That’s what everyone bills AI as being great at.
What most people don’t want from AI is, well, AI: They don’t want content that sounds like it was created by a beige conference room wearing a name tag. What they do want from AI is content that truly sounds like them. They want their ideas, tone, structure, humor, worldview, and especially their point of view to come through.
While it seems we’re drowning in AI-generated slop that sounds like AI rather than humans anymore, this is where AI can help.
Thing is, you (and your AI tool) need to understand what “voice” really menas.
Voice is not just word choice. It’s not just whether you use contractions, short sentences, or the occasional “Heck” or “Yeppers.” (Or full-on cursing, if you’re the Gary Vee-type.)
Voice is what you emphasize. It’s what you cut, how directly you speak to the reader. It comes from how you move from an idea to an example. It’s also what you refuse to sound like (hello, Gary Vee).
It’s also how you land the proverbial plane.
Many people get tripped up when they ask AI to write in their voice, but they have never clearly shown it what their voice is.
Or maybe more importantly, they have never shown the AI what their voice is not.
The First Draft Is Just the Starting Point
As with many of my off-the-cuff processes, what I went through was simple. I asked AI to draft my latest LinkedIn newsletter on the idea of being “professionally vague.”
My angle was that too many LinkedIn profiles sound polished but say almost nothing. They use fancy schmancy language, corporate fog, broad claims, and polished phrases that look professional but do not actually help the reader understand what the person does.
The first draft was solid and useful. Definitely had “me” in it, but wasn’t really me. So I edited it, then pasted my edited section back into the AI.
Didn’t stop there, though. I asked it specifically to compare what I had changed against what it had originally written. And with that comparison, I asked it to spell out what it had learned and how it would change its output for me going forward.
This is where I stumbled into some real magic. Because the point was not just to say, “Good job” or “Try again.” The point was to identify the pattern behind the edits.
For example, the AI version talked about the profile as a “positioning asset.” That is not wrong. It is a perfectly fine phrase.
But my edited version moved the section into more practical LinkedIn profile guidance. I wanted to explain what the headline and the About section should actually do. What the Featured section should support. And what the Experience section should show.
That difference matters.
While “positioning asset” sounds smart, “Your headline should explain who you help, how you help them, the services you provide, and the results you bring” is useful.
And useful almost always beats smart.
Radical, right?
Your Edits Are the Training Data
More people need to understand that sub-headline.
When you edit AI’s writing, you are creating incredibly valuable training material for your own future use. (You’re not training the model globally… let’s not get carried away here.)
What you’re actually creating is a reference point. You’re showing the AI what you prefer, what you changed, what you strengthened, what you removed, and what direction you want future drafts to take.
That matters because your edits reveal your writing instincts. In my case, the edits showed a few clear patterns. I wanted the writing to be more direct and instructional. The copy needed to move away from abstract positioning language and into practical guidance people could use on LinkedIn.
I wanted prospect-centered language; phrases like “your prospects,” “the results you bring,” and “the services you provide.”
I also wanted the profile advice connected to LinkedIn’s ability to understand what someone is relevant for. That’s a huge point for that particular article.
And I wanted the ending connected to a business outcome I actually care about: Earning more conversations without being salesy.
That’s voice. Not just the words; the worldview.
The Comparison Is Where the Learning Happens
The most useful step in my on-the-fly process wasn’t this “follow-on” prompt. It was the comparison.
While I’m going to work on this languaging for my future use, here’s what I initially told the AI to do:
Here’s my edited piece, based on what you’ve written. You need to learn from what I did and how I changed/edited your copy and apply that to new LinkedIn newsletters of mine going forward. Prove that you've learned by providing me with a comparison of changes. (I then pasted the copy here.)
I’m going to improve on this, but I got some really good results from just the above. For example, here’s the full section of the AI’s response to me from the “positioning asset” example I listed in the first section:
You shifted from “positioning asset” language to “profile as a working sales/conversation tool”
My draft leaned a little more abstract:
“Your Experience section should be more than a job history. It should show relevance, credibility, and outcomes.”
Your version says:
“Your Experience section should be more than a job history. It should show relevance, credibility, and your story.”
Lesson learned:
“Outcomes” is useful, but “your story” feels more human and more Bob. It reminds people their profile is not just a proof document. It’s also a trust-building document.
Throughout the article, when I asked AI to explain what I changed and what it should learn from those changes, the editing process became a feedback loop. I realized that this was huge as the AI became much more than just a content generator. It was truly learning what was important, and then applying what it learned to future drafts.
I didn’t realize it at the time, because I was in a time crunch to get that piece out. What I ended up thinking, though, is that this content feedback loop is what most people (unknowingly) skip.
They generate something. They dislike parts of it and make edits to others. They then either publish or discard it. But the next time, they start from scratch and wonder why the AI still does not “get it.”
Well, my friend, you actually didn’t give it anything to “get.” You didn’t teach it by showing the before-and-after. You didn’t explain why your edited version was better, and which patterns you want repeated. You didn’t point out the phrasing, rhythm, structure, or strategic emphasis you want it to remember.
By omitting all of this, you’re not really building a writing process. You are just asking for new drafts and hoping the robot gets less weird.
Teach AI What Your Voice Is Not
Your voice is not only defined by the phrases you use, the rhythm you like, or the topics you cover. It is also defined by what you reject.
Maybe you do not want to sound too polished, like a motivational speaker who got trapped inside a LinkedIn carousel. You don’t want to fake enthusiasm, spew jargon soup, generate corporate fog, or produce the kind of writing that technically says something while making no actual point.
That matters because AI often defaults to “professional” language, which becomes generic. That’s probably why we’re drowning in that AI slop I mentioned earlier… people aren’t doing the kind of training I’m discussing here, which leads to that default professional sound.
So part of teaching AI your voice is saying, “Not that:”
Not this phrasing.
Not this structure.
Not this level of polish.
Not this kind of CTA.
Not this fake-sounding enthusiasm.
Not this robotic conclusion.
That kind of feedback is not being picky. It is being specific. We also teach using what we call “prohibitions” in our CRISPY™ prompting framework. We even wrote a whole book about it, which you can find here. Prohibitions help AI sound more like you than like everyone else using AI.
Build Your Voice Loop
Here is the process I recommend:
Have AI draft something based on a clear prompt.
Edit the draft yourself. Not lightly. Actually make it better.
Paste your edited version back into the AI LLM.
Ask AI to compare the original with your edited version and ask it to identify the repeatable lessons.
Use those lessons in developing the prompt for your next piece of content, save them as part of your writing instructions, or ask your LLM to commit everything to memory.
It’s not complicated, but it does require the human to stay in the process. And that is the whole point, right?
The True Win Is Not One Better Article
The obvious benefit of this process is that you get a better piece of content, which is nice. But there’s a bigger benefit: You’re building a better system.
You’re creating a repeatable way to help AI understand your voice, style, audience, structure, point of view, and strategic intent. That’s where the leverage is… not in asking AI to write like you once did, but in teaching it to get closer over time.
And yes, you will still need to edit. Frankly, you should still need to (and want to!) edit. That is not a flaw in the process. That is the process. Your voice is not a template; your voice is a set of choices. The more clearly you show those choices, the better AI can support them, not replace them.
This is where AI truly becomes a useful tool in your sales toolbelt. Not because it magically becomes you.
But because you stayed human enough to teach it.
And always remember: Don’t be a salesy weirdo.
Earn those conversations.
NOTE: The hero image for this article was generated by AI (Gemini).
If your LinkedIn content is not moving your prospects from “I saw your post” to “I’m interested in speaking with you,” then you’re not aware of AWARE™.
While it’s not that simple to get people from that first point to the second one, the AWARE™ Framework gives you the progression you need to get your audience of prospects to the point where they want a conversation with you.
With AWARE™, you can drive your audience from recognition to curiosity, then to insight, next to a shift in perspective, and finally to action.
Get your FREE eBook on The AWARE™ Framework for content that converts… today!
