Why Your AI Content Doesn't Sound Like You (It's Not a Prompting Problem)

In June of 2023, Matt Goldman got a text from an agent he'd known for years. She wanted to know who was managing his social media — because it didn't sound like him. He told her it was him. She told him she'd been about to call out his social media manager for posting AI content.

That moment is where Creative Authority started.

The content wasn't bad. It was structured, readable, and professionally polished. By every measurable standard, the quality had gone up since he started using AI. But quality and authenticity are not the same thing — and someone who knew him well enough could feel the difference immediately.

In this episode, Matt breaks down the real reason AI content goes generic. It has nothing to do with your prompts, the tool you're using, or how much time you spend editing the output. The problem is structural — and it happens before you type a single word. Most agents open a new session and give their AI a task. They never give it a foundation. So the tool does what it always does when it has nothing specific to work from: it produces the average. A blurred composite of every real estate agent who has ever put anything on the internet.

Technically correct. Completely indistinct. And if you've spent years building a business on being specifically and recognizably you, indistinct might as well be invisible.

This episode explains why "write in my voice" doesn't work, what actually does, and the two foundational documents that change everything: the Mission Statement and the Canonical Voice Reference. These aren't style guides or tone settings. They're the documents that tell your AI who you are — how you explain trade-offs, what you'd never say in front of a client, the rhythm of how you communicate with people who are nervous, overwhelmed, or about to make one of the biggest financial decisions of their lives.

Without them, your AI defaults to its average voice — which is the average of everyone who has ever written anything professionally on the internet. With them, the output stops being generic and starts being yours.

If your content has started to feel off — if you read it back and it doesn't quite sound like you — this is the episode that names the actual problem. And it's fixable.

Transcript

SPEAKER_00 0:00

In June of 2023, I got a text from an agent that I've known for years, and she said, Who's doing your social media? And I said, Me. I'm running it all through ChatGPT. I went to a training. You love it. I'll send you the link. And she said, Yeah, I can tell. It doesn't sound like you. I was actually texting you to call out your social media manager. I want you to sit with that for a second because that hit different than I expected it to. This is someone who is my friend who knows me, not someone that I met at a conference or was like a client, even. This is someone who I had real conversations with who knows how I think and how I talk. And she read my content and she thought someone else wrote it. And she was right. Now here's the thing that I had to reckon with. I wasn't posting bad content. I was, it was structured. It was readable and professional. And by algorithmic measures on Instagram and Facebook, the quality had gone up since I started using AI. But quality and authenticity are not the same thing. And the content was better by those metrics, but it had completely stopped sounding like me. And I was too busy creating it and posting it and doing the real work in real estate of working with clients, then pay it to pay attention. And that part bothered me. It was so focused on the tool and what it could produce that I stopped asking whether what it was producing was actually mine. So I want to talk about what happened and more importantly, why it happens, because I am sure I'm not the only one. Welcome to the Creative Authority Podcast. I'm Matt Goldman. I am a real estate agent with 12 years of experience. I've built two real estate businesses across two different markets using AI, and I've done deep work in this space, and I'm helping agents now use AI in a way that amplifies you as opposed to quietly replacing you. And the reason I do this work is because I went through this thing that I just described. Like I lived it. So I figured better figure it out on this side. I lived it. So here's the problem. As cleanly as I can state it, every AI training that I've ever been to or attended leans heavily into the prompts. Here's the seven prompts to write a listing description, or here's the four prompts to write an email template, or the framework for a market update. Get all this stuff done in 30 seconds. And I'm not saying that those don't work. They do. They produce output, real, usable output, but they all skip a question that matters more than any of them. What is the AI actually working from? Because here's what AI training is doing right now it's teaching you to run before you even know where you're going. It hands you a tool and says, here, use this. And you use it and it produces something and you post it and it looks fine, but fine is not the same as you. And in a business built on trust and relationships, and the very specific way that you show up for your clients, fine is a liability. Let that sit. When you open a new chat and you type a prompt, the AI doesn't know who you are. It doesn't know what you believe about real estate. It doesn't know how you talk to a seller who just got a lowball offer. It doesn't know the specific way that you slow down a conversation when someone is feeling overwhelmed. It doesn't know what words you'd never use. It doesn't know the position and your beliefs that you've held for 15 years and why you hold them. It knows nothing about you specifically. So it does what it always does when it knows nothing about you specifically. It produces the average, the average real estate agent that it has ever read. Every generic listing description, every boilerplate follow-up email, every market update that you know starts with in today's real estate landscape, or it's a seller's market out there, all of this is blended together into something that technically is correct, but is vapid and completely indistinct. And that's what my content had become. Not bad, just indistinct. And when I built a business around being specifically and recognizably me, indistinct be it was as good as becoming invisible. I assumed the tool would just figure out my voice. I assumed if I used it enough or I gave it examples, I would eventually it would eventually start to sound like me. And it doesn't work that way. AI is not learning you, it reflects the quality and the specificity of what you put into it. Every single session, put in vague input, you're going to get vague output. Generic context, generic content. Every time, without exception, it is designed on averages. The agents who are getting consistent on brand output from their AI are not better at prompting.

unknown 5:32

Come on, Matt.

SPEAKER_00 5:33

They are not like technically more sophisticated. They did something before they ever wrote a prompt. They gave the tool a foundation to work from. And that foundation is made of two things. One, your mission statement. This is not a tagline, not your brokerage's value proposition, not everything I touch turns to sold, or helping you find your way home. This is a real working document, one that captures how you actually think, how you make decisions, how you serve specifically, and where your boundaries are. This is a document that answers what is AI permitted to do on my behalf and what must it never do. Because here's what happens without it AI will fill a vacuum with defaults. And its defaults are built for the broadest possible audience. The safest possible answer, the most agreeable possible take, it like softens your opinions, it avoids actual positions, and it hedges everything. That's one of the AI drifts of the week. If you're watching on Instagram, like it starts to soften all of your actual opinions by putting qualifiers in and saying, you should really consider, or if I were in your position, I would. And people are paying for your opinion. Any real estate agent who can't actually take authority and do the job and advise their clients with confidence is not worth paying. So clients can tell. They may not be able to name it or put their finger on it, but they can feel it. The conviction is gone, your specificity is gone. The thing that made your communication feel like it came from someone who actually knows something, that's gone. My agent friend knew that and she texted me too, and I'm forever grateful for her. So the second thing is the canonical voice reference. This is a document that captures how you actually communicate, written and spoken, the rhythm of how you explain something complicated to someone who's nervous, the words you'd never use in an actual client conversation, the way that you handle pushback, the way you open a difficult conversation before you have to say the hard part, that's not tone setting. That's not make this edgy or make this friendly. This is the real thing. This is the actual texture of how you show up for people, the architecture of your language. And the reason it has to be documented, the reason you can't just assume AI will pick it up is that your voice is not AI's training data. Your specific way of working is not publicly available anywhere. The only way that tool can know it is if you tell it explicitly before the session starts. So without this canonical voice reference document, AI defaults to the average voice the same way that it defaults to the average content. And the average voice is like a blurred composite of every professional who's ever typed into it or put anything on the internet, period. That's just what my content sounded like. It wasn't bad, wasn't offensive, it wasn't wrong. 82% of the time, it wasn't wrong. It just wasn't me. Your AI sounds generic because you told it nothing real about yourself. That's the whole thing. The entire problem in one sentence. It's not the tool, the tool's fine. It's that most agents open a new chat, type a request, and expect that tool or prompt, and expect that tool to know something in their market, in their personality, and their 15 years of judgment and the very specific way they've earned the trust of people in their business. And it doesn't know any of that. Like it can't unless you put it in there. Now I want to be clear about something because I think this can get misunderstood. I'm not telling you that AI doesn't work. It works. I use it every day. I'm telling you that AI without a foundation is like hiring an assistant and never briefing them. Like they'll do the work. Hopefully, they'll be competent and they'll do it in their own way, but that's going to be average. They're going to do it safe. They're going to make it so they don't get fired, basically. And that's not your way. If you're a top-producing agent, if you're a veteran agent, you have beliefs, you have things that you can say. You didn't hire an assistant, you got an AI. So make your AI do the thing that you want it to do. That foundation is the briefing. It's the thing that tells the tool who you are, what you stand for, and how you're going to show up for people. And once you have it, everything changes. The output stops being generic and starts being yours. Not perfect, yours, which is pretty close to perfect. But that's the whole point. Why else are you using AI? Here's what I want you to take from this. If your content has started to feel a little off, if you read it back and it doesn't quite sound like you, that's not a prompting problem. That's a foundation problem. And it's fixable. The work of building that foundation is not complicated. It does not require a conference in Toledo, Ohio. It doesn't require you to become a different kind of agent or learn a new set of technical skills. It requires you to write down what you already know. What you believe, who you serve, how you talk, what you'd never say, the positions that you've held over your career that you've never actually put down on paper. That's the work that creative authority is built around. The mission statement, the canonical voice reference, the research lens, the marketing statement. That foundation is what makes everything else work. And if you've ever spent money on AI training that hands you prompts with no foundation under them and wonder why the output still doesn't sound like you, that's why. And that's what we fix. So the link is in the description if you want to know more. I hope this triggered something for you, like it did for me three years ago. I'll see you in the next one.