Why Your AI Real Estate Copy Doesn’t Sound Like You

If you ask ChatGPT to write a listing description, you’ll get something competent, grammatical, and completely anonymous. If you ask it again next week for a different property, you’ll get a draft with the same rhythm, the same adjectives, and the same shape. It probably sounds like every other agent’s copy because it is every other agent’s copy.

The usual conclusion is that the tool isn’t good enough yet, but that’s not what’s actually happening.

An AI tool that knows nothing about who’s speaking writes the average of everything it has read. For real estate, that average is decades of listing descriptions and neighborhood write-ups, most of which were written fast and none of which were written to stand out.

So, it hands you the median of all real estate copy. And the median of all real estate copy is exactly the thing people mean when they say AI writing sounds like AI writing. You got generic output because you gave it nothing to be specific about.

Prompt tinkering just gets you a different flavor of average. The fix is an input that –  for almost every agent – doesn’t exist in writing anywhere: a description of how you sound.

Think about what you actually have written down. What you almost certainly don’t have is a page that says what your copy should sound like, who you’re talking to, and, most importantly, what you’d never want to sound like. It’s just living in your head, where the AI can’t reach it.

If you write it down right now, your next draft will improve immediately.

Answer these three questions to start

  1. What would you never want to sound like? This is the single highest-value question in the exercise, and almost nobody has answered it.
  2. What are the ten phrases you’re tired of seeing in your market? Write them out as a banned list.
  3. What can you actually claim verifiably? Your years in the business, neighborhoods you truly know, and your real specialties. 

Paste those answers at the top of your next chat before you ask for anything. Everything below is just making it sharper.

Why the “never” list does the most work

Every agent’s always list is basically identical: warm, local, trustworthy, and knowledgeable. Those words steer nothing because nobody claims the opposite.

What separates your voice from the agent down the street is what it refuses to be. Here’s a sample from a fictional suburban brokerage, Maple Realty:

Never: hyped, urgent, or pushy. Overly polished or corporate. Talking down to first-time buyers asking real questions. Making lifestyle promises, or describing who a home or a town “suits.” Use of slang that’ll read as dated in a year.

You learn more about that brand from those five lines than from any tagline it could write and the AI can act on every one of them.

Then, give it the actual banned words:

“perfect for families,” “great for kids,” “make memories that last,” “won’t last,” “act fast,” “priced to sell,” “your trusted partner,” “we go above and beyond.”

Naming them is what keeps them out. “Avoid clichés” does nothing, because the model doesn’t think it’s writing one.

What makes a voice click

Every trait you want has a nearby version that ruins it. If you name the near miss, you’ve said something genuinely useful:

  • Warm does not mean sickly sweet or over-the-top.
  • Approachable does not mean dumbed-down.
  • Confident does not mean pushy.
  • Local does not mean small-town-insular.
  • Reassuring does not mean patronizing.

Copy doesn’t usually fail by missing the target; it fails by overshooting into the bad version right next door (polished slides into corporate and friendly slides into gushing). The failure is always one step too far, so that’s where you put the fence.

Don’t skip Fair Housing

A voice instruction like “sound warm and local” can become permission to describe who a home is for, and that’s the road straight into Fair Housing trouble. For agents, it’s the easiest mistake to make because the whole habit of the business leans on “great for families,” “good schools,” “safe neighborhood.” Every one of those describes people, not property.

Use this rule on top of everything else:

Describe the property, describe the geography, the commute, the building, and the amenities. Never describe who belongs there.

Any tool like this helps you write, but it’s not a compliance review. What you publish is on you to review and approve.

If you’d rather not stare at a blank page, we built a free custom GPT that runs the interview for you. It asks how you work, who you serve, what you sound like, and what you refuse to sound like, then writes the finished one-page blueprint. It shows you a plain-language draft first so you can correct it before it’s final.

The interview is half the value. Most of these answers have never been said out loud, and being asked what you’d never want to sound like is, for a lot of agents, the first time they’ve had to think about it or decide.

What you walk out with belongs to you; it’s not locked to anything of ours, and it doesn’t expire when you switch models.

Want the link? Request it here, and we’ll send it over.

How strong is the page it lands on?

A voice blueprint fixes the words. But words only work if the page they live on can actually be seen.

The pages that should pull the most search and AI traffic –  your listings, your search results, your neighborhood pages – are often the ones AI systems struggle to read, because the good stuff hides behind interactive filters and maps until a browser loads it. Perfect copy on a page a crawler can’t fully see does close to nothing.

So, once your drafts sound like you, ask the harder question: can the systems buyers now search with actually read the page your copy is sitting on? We dug into why your best pages are the hardest to surface and how it gets solved, here.