It's like asking someone you've never met to write your wedding speech.
The output isn't bad. It's grammatically correct, logically structured, and completely empty, like it was written by someone who has read everything and experienced nothing. And the frustrating part is you can't point to a specific sentence and say that's the problem. The whole thing just doesn't sound like you.
That's not a prompt issue. That's a context issue. Almost nobody figures out the difference.
What "no voice" actually sounds like
You know the feeling when you ask for a caption, a bio, an email, and what comes back is technically fine but all the wrong words. It's enthusiastic where you'd be dry, formal where you'd be direct, using phrases like "I'm thrilled to share" when you'd just say "this is out." Three exclamation points in a paragraph where you'd use zero.
The AI isn't broken. It just doesn't know you. It knows language, and those are different things.
What it's doing by default is averaging. It's pulling from the center of everything it's ever seen written in that format, by that type of person, in that context, and the result is something technically correct and completely generic. It sounds like everyone and no one at the same time. That's the voice problem, and it doesn't go away just because you rewrite the prompt.
Why it keeps happening
The default move is to give AI a task when what it actually needs is context.
"Write me a LinkedIn post about my new product launch." That's a task, so the AI fills it with the average of every LinkedIn post it's ever processed about a product launch, which is why it sounds like every LinkedIn post about a product launch. The ones you scroll past without reading.
What you actually needed to give it was who you are, how you talk, what you'd never say, and what this launch actually means to you. Not the assignment. The texture. The people whose AI output sounds like them figured this out, and it's not because they found a better prompt template. They're giving better context, and there's a meaningful difference between telling AI what to do and giving it enough to actually work with.
What you're actually losing
When your AI sounds like nobody, people feel it before they can name it. They don't read it and think "this is AI-generated." They read it and think "this doesn't sound like him," or worse, they just scroll past. The content is fine, it just doesn't stop them, because it doesn't sound like a person they want to follow.
Your voice is what makes your audience feel like they're talking to you and not a content machine, what makes someone recognize your writing before they even see your name on it. Lose that and you've lost the thing that was actually working.
AI is a tool. The voice is still yours to protect.
How to actually fix it
Feed it you. Not a description of you, but actual evidence of you. Real sentences you've written, captions that landed, emails you'd send to a friend, notes you typed into your phone at 11pm. The AI isn't going to reverse-engineer your voice from adjectives like "authentic, direct, conversational." It needs examples, so give it the real thing.
Beyond that, give it your constraints, because what you'd never say matters as much as what you would. "Don't use exclamation points, don't say 'dive in' or 'game-changer,' don't wrap up with a question asking people to comment." Negative constraints cut the generic faster than positive prompts add specificity.
Then push back on the first draft the way you'd push back on a person. "That reads like a press release, make it tighter, drop the last two sentences, one person not a crowd." That kind of directional feedback is more useful than rewriting the whole prompt from scratch. Iterate with direction, not desperation, and it gets better fast.
What I built
Here's what I did with all of this. I documented my own voice, my patterns, my hard constraints, the phrases I'd never use, and built it into a reference system. Then I built 56 prompt transformations: real before and afters, stiff to natural, robotic to human, generic to actually-sounds-like-you, covering the situations where voice matters most: writing online, selling something, thinking through a problem, everyday use.
It's called Human Mode and it's $12. Not a course, not a framework with six pillars. A practical reference you use when your AI output sounds like nobody, because you want it to sound like you.
Filed under AI — part of the OWNWARD series on tools and systems for people rebuilding in public.