I’ve been reading content from one of my favorite entrepreneurs since 2019. They’ve written books, solved gigantic problems, worked with household names, and they’re easily one of the top experts in their field.
I’m a big, big fan of this person's work, and…
I just unsubscribed from their emails.
I’ll hide some details to protect the guilty, but it’s so ridiculously obvious that they’re using AI to create their content and they’ve lost the parts that used to make it great. And I can say with 100% certainty that they had previously defined their voice to an exceptional level.
(I know their former editors and copywriters and can confirm they have a very thorough voice guide.)
Prior to 2024, their content was some of the most resonant information you could find on their subject. This person wrote most of it themselves, yet used copywriters for some marketing materials or scripts depending on the medium. It was almost always a consistent voice and met a very high standard.
Yet since their wider implementation of AI, that standard dropped. And it feels like they're phoning it in.
So let’s talk about what went wrong…
What Happened: The Death Of A Great Voice
Lemme start by saying that I don’t have a problem with AI. I use it for tasks all the time — and if it helps you write or think, that’s great. But please stop fooling yourself that nobody will notice.
We will.
Some of us care more than others, but your audience is smart and deserves your respect. They deserve you.
But how do you do that? How do you humanize your AI content? How do you give yourself to your audience and put yourself in your content — even if you didn’t write it?
The answer is: you’ve got to start with what it wrote right.
Because AI is great at doing things correctly. It sanitizes mistakes, sands off edges, and softens the hard parts to make your writing clean.
It does things “by the book.”
But in doing that, it loses your style.
And to quote Neil Gaiman:
“Style is the stuff you get wrong.”
See, we don’t always follow the rules. We create our own writing styles — and in doing so, we get things “wrong.” It turns out that the stuff you get wrong is the humanity.
Look, this ain’t something you fix with a prompt. It’s not something you tweak here and there, and it’s not something AI will replicate. Because AI’s very nature is to base its content on a world's knowledge rather than yours. Which means it’s great at knowing stuff, but terrible at knowing you.
And it tries to “correct” you just as much as it tries to mirror you.
AI writes to appease an English teacher — grammatically correct, factually accurate, and observant of all the rules. The problem is, they’re not your rules.
Human isn’t a writing style.
Which means you can’t vibe code a voice.
Asking Claude or Chat to “make you sound human” means that it’s going to make you sound like the wrong human. It’s gonna sound like the masses.
Ya gotta make it sound like you. But in order to do that, ya gotta know what you sound like.
Common Mistakes: What most people try
I already said that prompting won’t fix this. It’s not something you nudge into place, and unless you’re writing something completely basic and objective, AI needs more than most people give it to capture their approach. Yet still, the best results I’ve seen from any AI platform come from knock-down, drag-out fights and arguments where you bring the fire and feist and rile up your opinions.
Basically, you yell at Claude until you finally tell it what you want to say, which is basically the same as writing it yourself... just more stressful.
You can absolutely make those passionate parts a piece of your voice guide afterward — and I think you should. Because, aside from your punctuation preferences, that’s the stuff that we wanna hear. That’s what makes you you.
Without it, most voice instruction sounds completely vanilla. It’ll default to the average “human” (because we literally tell it to “write human”) and water you down to three tones: “friendly, casual, and bold.”
The Trifecta Of Nothingness.
(You can hear my rant about this on The Message-Market Fit Podcast)
I’ve already talked about the problem with “human.” (There are 8.5 billion humans on the planet. Choose one.) So let’s talk about the three other useless tones:
Friendly:
Think about your three best friends. Do they all sound the same? No, and that’s probably what you love about them. They may share similar personality traits, and that’s an important distinction.
Friendly isn’t a voice. It’s a personality.
It’s your actions and motivations. It’s more about what you do than what you say… and it can come across countless ways. Voice is just a tiny piece of your personality and without specific direction and documentation, saying your voice is friendly doesn’t mean bupkis.
- Hi
- Hey
- Hello
- Greetings
- Saltuations
- Wassssuuuuuuuuuup
- Welcome, dear heart
- ‘Sup nerd!
…Those are all “friendly” depending on your relationship with the audience… but clearly, they’re very different voices.
Casual:
You can define my style as casual ‘cuz I sometimes drop my G’s, abbreviate words, use PG-13 language, and make random asides and references. (I’m a big fan of butting into my own stuff.) Yet to someone else, “casual” might mean using exclamation points or emojis, whereas I use neither.
And lemme ask you if the idea of writing “lemme ask you” suits your style, or does it give you the vapors?
Casualness is a spectrum — and without further definition, it ain’t gonna get you what you want. Plus, AI has its own levels of casualness, and the chances of it matching yours are pretty dang slim.
(See, even there I could have said “slimmer than a pine needle, slimmer than a Slim Jim, less than 3.1%, slimmer than your old boss on Ozempic, lower than your chances of making the Kessel Run in 12 parsecs… or simply, “slim.” Those are all parts of voice.)
Bold:
First off… bold ain’t a brand voice. It’s a font. OK, technically a type face, but just like friendly, bold defines actions, not writing. You can say you write confidently and avoid hedge words. You can write declarative statements and remove questions. You can write short.
Those directives sound bold, but ya gotta back it up with what you say.
“I am timid” is technically a bold statement because it follows the rules of the above.
But if you wanna talk bold, ya gotta act bold.
- Offer a 100% guarantee
- Call your shot
- Change production to a costlier-but-more-ethical supplier
(It can, but ya gotta do it right)
When it comes to being bold, voice is the supporting act.
Bold, friendly, casual, and human are the wrong adjectives. They circle the problem, but don’t fix what’s missing. And they might get your AI output a teensy bit closer, but they don’t really solve anything
Because the thing that’s missing — the REAL thing missing in AI copy… is cost.
AI never has to pay a price for what it says, so it never takes a risk.
Good writing does.
The REAL fix: Six tips for bringing back humanity into AI writing
Look, I know that “chapter marker” subhead right above this line is an AI tell… but the truth is, that’s me trying to appease the algorithms and get this to show up in searches. And I hate that I have to do this, but here we are…
Either way, the real way to get yourself back into your writing is, well… to put yourself back in the writing.
Writing or editing with AI strips out the “risky” things. It trims out all the “unnecessary” parts that are the exact things that make it you. Those “extras” are the voice. It’s what gives it the you-ity-ness and grit.
You make “mistakes” and say stuff that safe writing doesn’t. Those are your fingerprints that AI scrubs off.
So, how d’ya put yourself back into writing? I’ll yell ya, baby bird:
Human writing contains consequences.
Write something costly:
Give your content an admission that puts something at risk. Your authority, your stance, your opinion, your take… it doesn’t have to be world shattering, but get off the damn fence and make a choice. And I’m not talking about piddly lil’ junk like if a hot dog is a sandwich… I’m talking about an actual thing your industry cares about that might split some of the readership.
You don’t have to alienate people — and it doesn’t have to be a socially polarizing topic — but real human writing isn’t afraid to show its opinion. (Again, this doesn’t have to be a political, social, or ethical topic… those only have to be part of your business if you want them to be. But for glob’s sake, take a stance on something!)
If you want to stand out, you have to stand up.
“To stand for something takes more courage than to just be critical”
- Shepard Fairey
Add contempt:
Tell people the rationale behind your decision and explain “your enemy.” For me, I think most of the AI writing advice is bullshit and it's written by people who don’t really know what good writing is. I hate that we’re losing ourselves to machines and I think most “experts” out there are opportunists selling trivial products based on false promises and fooling themselves that their AI content is “good.” Or even “good enough.”
I’m not saying you need to go full manifesto here (though, I do love me a good manifesto) but a public declaration of intent and what you’re for and against can move men, women, and mountains.
Name Names:
AI looooooves to attribute vague advice. “People say” and “experts claim” type of language. But most of the best stuff comes from a real person.
Now, you don’t have to always point fingers if it’s a negative thing… for instance, I talked about “an entrepreneur” in the intro to this piece who is very real. But I’m not gonna say their name or even allude to their gender or industry out of respect. Also, you’ve likely seen someone else do the exact same thing, so I’m focusing on the act more than their name. However, I will often tell you my opinions on non-personal brands, software, products, and specific practices.
There are kind and unkind ways of doing this and you’ll likely know what's right for you. But the more specific you can be, the more “human” it will feel.
Use idioms, slang, and weirdness:
AI digs using cliches that don’t look like cliches. At the time of this writing, it’s relying heavily on “3 _____ in a trench coat” framing — which is funny, but now feels generic.
Instead, use your line. Movie and TV references, song lyrics, oddball analogies, made up words, obsessions, opinions, preferences, taboos, quirks… they’re all great. In this piece alone I’ve made Star Wars references, said “youityness”, thanked glob, and called you “Baby bird.”
Those are Justin-isms. Work in your quirks when natural. AI will pick up on it and start to over-use ‘em, but at least it’s trying. Figure out your own system for knowing what’s what.
(Star Wars, for me, is only the original three. Baby bird is when I address a common misstep, and glob is because we all have our different beliefs and I appreciate that.)
Break cadence:
Cadence is the rhythm of your writing. AI likes short n’ choppy ‘cuz it’s easier to read. I write like that too. And then I get on a tear or go on a side rant that extends way beyond the average sentence length of fifteen words, and may even double or triple it until it starts to feel unruly and you’re not sure if you want to speed up or slow down and it sorta breaks your brain (because it’s also breaking my brain) and even though every online editor like Hemingway or Verbatim would tell you trim it, you’re gonna leave it anyway, by golly, because this is the way you’d say it.
Like that.
That’s an example of non-AI writing in a nutshell. (That cuckoo bananas sentence had 93 words, BTW.)
Keep the ending open:
Claude and Chat like to tidy up topics with a neat lil’ bow. They want you to understand the takeaway and make sure you walk away with the payoff, even if that means hammering you over the head with it.
But some endings are up to you. Sometimes the best pieces make you think rather than act… and leave you to choose for yourself.
As much as I like direction, letting the reader choose their own destiny is often a kindness. Because sometimes there’s a best way… and sometimes there’s just a best-for-you way. Guiding a reader to a decision will often make you more memorable (and more human helpful) than telling someone directly.
Wanna know the worst part about this list?
These are the exact things that AI removes when you ask it for help.
So, how do you get AI to actually sound more like you?
The real fix isn’t a prompt. There’s no hack here and there’s no secret. But you absolutely can decide what flags you’re willing to plant, what lines in the sand you’re willing to draw, and what side you’re on.
Build your values alongside your voice and let them work in harmony.
That’ll add the depth you’re missing.
You’ve probably noticed that AI loves to talk about “gaps.” It uses a lot of “X but not Y” negative parallelism and focuses on borders rather than the body. Meaning, it’s good at talking about the edges of an argument, but never the center.
That’s because it has a better idea of what things are not than what they are. It doesn’t know where you stand, so it predicts where you don't. And it’s trying to dial you into a general area through subtraction, but leaves the range too wide to have much meaning. Which is why everything feels “meh.”
It’s like a map telling you what city you’re in, but not what street.
To teach it, you can fight with it during each conversation and write down instructions for future use. That’ll help you get clearer and slowly create your own voice guide. And you can feed it some of your more personal writing — like a journal entry or some of your most popular emails. Also share stuff where you took a risk. Give it an email you were scared to send, and it’ll help it see your stance.
Start a new chat and ask AI to document patterns within your personal content — and add it to its memory. (Remember, each chat is its own sandbox and it won’t necessarily carry over to your AI’s main brain. You can apply rules to projects or your AI profile itself, but you have to instruct it to do so. The exact mechanics vary by tool. If you’re not sure what to do, you can literally just ask it.)
AI doesn’t need more instruction about how to sound like you. It needs guidance about what you’re willing to say. Because unless it understands the decision making that produces your voice, it’s gonna lean toward a surface-level imitation of you. And it’s gonna play it safe.
If you’re working with a traditional or corporate brand, your content is less likely to have risk, but you’ll likely have some stuff with a stance. Or you can stick to AI's default measures, but you’ll need to be satisfied with sounding like B-roll: safe, exportable, and average.
Even if you don’t put these more emotional examples into your AI, put them into your brain. Read your realest pieces of content and ask:
- What rules did I set for myself when being vulnerable, excited or agitated?
- What lines was I willing to cross and which were off-limits?
- What language and cadence patterns changed?
Find your recurring risks.
AI is pretty good at identifying these differences IF it understands what it’s looking for. The real trick is training yourself to identify what’s different too. Then you can put rules around them for yourself, and…
Then you can do it on purpose.
And that’s when the real humanity begins.
The goal isn’t to make AI sound more human. It’s to make it sound like YOU.
Even if you know your voice well, this stuff is five A’s hard to do on your own. It’s haaaaard. And it’s almost impossible to see what you do differently from others because it feels natural to you. It’s not until you see it done “wrong” that you even knew this was a thing.
And then you’ve gotta decide which way is “right.” Or at least, right for you. (Because those two things are very often different.)
The real trick to getting your voice right is understanding exactly why it feels wrong — and knowing your patterns on a deeper level than you likely do right now.
I’m not talking about the generic stuff. I’m talking about the meat on your voice’s bones. The good stuff that makes you identifiable.
Your voice is in the details that most AI can’t see and most people don’t know exists.
I’ve analyzed more than 549 different voices from people and brands across countless industries, and sometimes it’s still tricky for me. But I’ve got a process that brings it to life and documents it with very little effort on your part. Only takes two hours on your end. All you do is show up and point.
Voice Finder Sessions are where we dive deep into your brand voice and collect all the specific elements of your true style (the one you actually have, not the one you want) — and put it all on paper.
From there, you can lean into it, evolve it, or adjust your voice to make it stronger than oak — and make it sound exactly like you.
Because you are the only human that matters.
Oh, and if you're new around here, I'm Justin Blackman and I talk a lot about brand voice and copywriting. If you dig this kinda stuff, you should totes get my emails.
