Standing Out When Everyone's Content Is AI
98% of brokerages now use AI — the tools stopped being the edge. Every feed in your market is filling with the same fluent, frictionless, forgettable copy. As Ryan Serhant put it: if we're all using AI and we all have the same expertise, "it's the game of attention." The winners aren't the agents with the best model. They're the agents whose content could only have come from one person. This session builds that machine: teach AI your actual voice from your actual writing, filter the slop out of everything it drafts, and scale a content presence that sounds so much like you that clients say "I felt like I knew you already" at the first appointment.
Four numbers that explain why sameness is the new invisibility
A voice system only works if every word of it stays true
This whole session is about authenticity at scale — which means the fastest way to destroy its value is to fake any part of it. AI amplifies your voice; it must never invent your life. Six rules, wired in before the first prompt:
| Rule | What It Means in Practice |
|---|---|
| Your stories must be true | AI scales your voice — it never fabricates your experiences. Every anecdote you publish happened. The moment a neighbor spots an invented story, every authentic post you ever wrote gets re-read as fake. |
| Your opinions are yours | AI sharpens takes you actually hold; it never generates a position for you to wear. If you wouldn't defend it out loud at a closing table, it doesn't get published under your name. |
| Client stories: permission + anonymity | Real stories often involve real clients. Get permission, strip identifying details, or generalize ("a seller I worked with this spring"). Confidentiality outranks content. |
| Fair housing applies to personality too | A "voicey" post is still a regulated post. Humor, opinions, and neighborhood color must describe property, price, and lifestyle — never who "belongs" where. Add the compliance line to every standing prompt. |
| Follow your brokerage's AI policy | Know your brokerage's and state's rules on AI-assisted content and disclosure — and if a client directly asks whether you wrote something by hand, tell the truth: "I write with AI, but every word is mine before it ships." |
| Amplify, don't act | Voice work turns YOU up — it doesn't build a character. Exaggerating a persona you can't sustain in person creates a gap the first phone call exposes. The test is always: would the in-person you recognize the online you? |
The point, stated plainly: in a feed full of synthetic voices, verified humanity is the scarcest asset in your market. These rules are what keep yours verifiable.
Teach AI to sound like you — from writing you already have
Generic output isn't an AI problem — it's an input problem. Ask a model for "an engaging post about spring listings" and you get the statistical average of every agent post ever written, because that's exactly what you asked for. But these models are extraordinary mimics: show them ten samples of how you actually write — the emails clients reply to, the texts that get deals unstuck, the one post that popped — and they can map your fingerprint: sentence rhythm, favorite phrases, how you open, how you land, what you never say.
The Voice Print captures that map in a Voice Card — a compact document defining your tone, cadence, vocabulary, opinions, and (just as important) your anti-style: the words and moves that instantly don't sound like you. Paste it into any chat, load it into ChatGPT custom instructions or a Claude Project once, and every draft starts 80% you instead of 0% you. That's the difference between editing your own voice and fighting someone else's.
Run in order, once — the output compounds forever
Use your best tool (ChatGPT, Claude, or Gemini) and keep the whole sequence in ONE chat — each prompt builds on the last.
I'm a real estate agent and I want you to learn my authentic writing voice. Below are [N] samples of my real writing — emails, texts, and posts I actually sent [PASTE 10+ SAMPLES]. Analyze them like a linguist: my tone and energy level, sentence length and rhythm, how I open and close, signature words and phrases I repeat, punctuation and emoji habits, how formal or casual I run, how I handle humor, and what I talk about when nobody assigns me a topic. Then summarize my voice in 10 specific, quotable observations — with an example from my samples for each. Don't flatter me; be precise.
Now interview me to capture what writing samples can't show. Ask me 12 questions, ONE at a time, covering: my three strongest opinions about my market and this business, the advice I give every client, my pet peeves about how agents talk, phrases I say out loud every week, the stories I find myself retelling, what I want people to feel after reading my content, and how my closest client would describe the way I talk. Push for specifics — if I give you a vague answer, ask the follow-up. At the end, summarize what the interview adds to my voice profile.
Based on my samples and interview, build my ANTI-STYLE list — the negative space that makes my voice recognizable. Include: 15 words and phrases that appear constantly in AI-generated real-estate content but never in mine (start with "nestled," "boasts," "stunning," "dream home," "look no further," and find the rest from contrast with my samples), sentence patterns I never use, an emoji policy based on my actual habits, hype levels I never reach, and openers/closers that would instantly feel off-brand on my feed. Format it as a hard ban list I can attach to any content prompt.
Compile everything — the analysis, the interview, the anti-style list — into my VOICE CARD: a document under 350 words that I can paste at the top of any AI chat to make it write as me. Structure: WHO I AM (one line), TONE (3 precise adjectives with a sentence on each), RHYTHM (sentence length, pacing, paragraph habits), SIGNATURE MOVES (my openers, closers, and repeated phrases — quote them), OPINIONS I HOLD (so content can take my positions), HARD BANS (the anti-style list, condensed), and A CALIBRATION SAMPLE (my single most representative paragraph, quoted in full). Make every line specific enough that a stranger could impersonate me convincingly.
Calibration test. Write the same short post about [TOPIC — e.g., what buyers get wrong about our fall market] three ways: Version A with no voice guidance at all, Version B using only my Voice Card, Version C using my Voice Card plus mimicking my calibration sample directly. Label them randomly as 1, 2, 3 and don't tell me which is which. I'll pick the one that sounds most like me and tell you what's off about the others. Use my feedback to update the Voice Card — specifically the parts that let the wrong version fool me.
Turn my finished Voice Card into three deployment formats: 1) a version trimmed to fit ChatGPT's custom instructions fields (with exact instructions for where to paste it), 2) a system-prompt version for a Claude Project or custom GPT dedicated to my content, and 3) a one-paragraph "quick loader" for when I'm in a random chat and just need 70% of my voice fast. Then give me the maintenance rule: what kinds of new writing to add to the card, and a calendar reminder text for re-running the blind taste test quarterly.
The quality gate between "AI drafted it" and "I published it"
Even with a great Voice Card, raw AI output regresses to the mean — the model's gravity always pulls toward the average post. The Slop Filter is the standing 90-second pass that pulls it back: detect every phrase that could appear under any agent's name, rewrite contrastively (showing AI the generic version and demanding distance from it), and apply the read-aloud test — if you wouldn't say the sentence across a kitchen table, it doesn't ship.
Then the move no model can replicate: story injection. AI knows the market; only you know what happened to you on Tuesday. The showing where the buyer cried in the driveway. The seller who almost took the hallucinated Zestimate to the bank. Sixty seconds of your real week, fused into the draft, is the one ingredient that makes content literally impossible to duplicate — because nobody else lived it. Slop is what everyone can generate. Your Tuesday is what nobody else can.
| Pass | The Question | What Happens |
|---|---|---|
| 1 · DETECT | "Could any agent in my zip have posted this?" | AI flags every generic phrase and scores the draft's genericness 1–10 |
| 2 · CONTRAST | "What would the average post say — and how do I not?" | The contrastive rewrite: same message, maximum distance from the average |
| 3 · INJECT | "What did I actually see this week that proves this point?" | Your real story fuses with the market point — now it's unduplicatable |
| 4 · READ ALOUD | "Would I say this sentence to a human at a kitchen table?" | Every sentence you wouldn't speak gets rewritten as you'd say it |
Your Voice Card loads first — then these run on every draft
Every prompt here assumes your Voice Card is already in the chat. That's the point of System 1.
Here's a draft post [PASTE]. Audit it for slop: highlight every word, phrase, and sentence that could appear in ANY real-estate agent's content — the interchangeable parts. Score the draft 1–10 on genericness (10 = pure wallpaper). Then tell me: which single sentence is most mine, which is most nobody's, and what percentage of this post would survive if a reader covered my name and had to guess who wrote it.
Take this message I want to communicate: [POINT]. First, write the AVERAGE version — exactly what a thousand agents' AI would produce for this topic. Then write MY version using my Voice Card, with one instruction dominating: maximum distance from the average version while keeping the message true. Different opener, different structure, different vocabulary, my rhythm. Show both side by side, then list the five specific choices that separate mine from the average — so I learn my own fingerprint.
Here's the market point I want to make: [POINT]. And here's what actually happened in my business this week, raw and unpolished: [TELL THE STORY IN 3–5 SENTENCES — names anonymized]. Fuse them: the story carries the post, the market point lands as its lesson. Keep my real spoken phrases from the story intact — don't smooth them into writing-speak. My voice, under 150 words, no moral-of-the-story cheese, fair-housing compliant. The reader should finish it knowing something happened to a real person this week.
Run the kitchen-table test on this draft [PASTE]. Go sentence by sentence and flag every line I would never actually SAY out loud to a client sitting across from me — writing-voice constructions, LinkedIn-isms, anything performative. For each flagged sentence, rewrite it as I'd speak it, using the rhythm and phrases in my Voice Card. Keep contractions, keep the fragments if I'd use them. When you're done, the whole post should read like a transcription of me talking on a good day.
Rewrite this listing description [PASTE] with every real-estate cliché banned: no "nestled," "boasts," "stunning," "oasis," "dream home," "look no further," "won't last long," or anything from my anti-style list. Instead: lead with the one detail about this home a visitor actually remembers [TELL ME WHAT IT IS], write like I'm describing it to a friend on the phone, and let specifics do the selling — the afternoon light, the walk to coffee, the thing the sellers will miss most [IF KNOWN]. My voice, fair-housing compliant, under 120 words. Every listing in my market sounds the same; this one shouldn't.
Monthly check-up. Here are my last 10 published posts [PASTE] and my Voice Card. Compare them honestly: where is my published content drifting back toward generic — which banned phrases crept in, which posts fail the cover-the-name test, and is my voice getting stronger or weaker over time? Rank the 10 posts from most-me to least-me. For the bottom three, diagnose exactly what went wrong (skipped the filter? no story? rushed?). Give me one specific correction to apply next month, and note anything new I'm doing well that should be ADDED to the Voice Card.
One voice memo. Four pillars. A week of content that's unmistakably yours.
Serhant's math is brutal and correct: when every agent has the same AI and the same information, "it's the game of attention" — and attention compounds to whoever shows up most consistently while sounding least like everyone else. The problem was never that you had nothing to say. It's that typing is where your voice goes to die. So the engine starts where your voice actually lives: talking.
Every Monday, five minutes, phone voice memo: what happened last week, what you're seeing in the market, what's annoying you, what deal taught you something. Transcribe it, hand it to AI with your Voice Card, and extract the week's content — with your actual spoken phrases preserved, because those ARE the voice. Then spread it across the four pillars that make a feed feel like a person: local authority (what's really happening), honest opinion (what you think about it), personal story (what happened to you), and proof of work (what you did about it). Fifteen minutes of talking. A week of content nobody in your market can duplicate — because it's made of your week.
| Pillar | What It Sounds Like | Why It Wins Attention |
|---|---|---|
| 🏘️ Local Authority | "Here's what actually closed on Maple last month — and why the number surprised me" | Real data + your read = the post AI-only agents can't fake |
| 🗣️ Honest Opinion | "Unpopular take: open houses aren't for selling YOUR house" | Positions are memorable; summaries are wallpaper — opinions start conversations |
| 📖 Personal Story | "My buyer cried in a driveway Tuesday. Here's what the tears were actually about" | Stories are the only fully AI-proof content — nobody else lived them |
| 🧾 Proof of Work | "What 14 days of prep looked like before this listing went live" | Receipts beat claims — showing the work sells the next client |
The weekly ritual, from memo to published week
Voice Card loaded first, always. Prompt 1 runs every Monday; the rest run as needed through the week.
Here's the transcript of my Monday voice memo about last week [PASTE TRANSCRIPT]. Extract every usable content seed and sort them into my four pillars: local authority, honest opinion, personal story, proof of work. For each seed: the hook (using MY actual spoken words from the transcript wherever possible — don't polish them into writing-speak), the pillar, the platform it fits best, and what's missing (a number, a photo, a permission) before it can ship. Flag the single strongest seed of the week and tell me why it's the one.
From this week's seeds [PASTE PROMPT 1 OUTPUT], draft my content week: 4 posts, one per pillar, each in my voice per my Voice Card. Vary the energy — not everything is a lesson; some posts are just me noticing something. Each post: hook in the first line, under 150 words, ends in a way that invites replies without engagement-bait ("comment YES below" is banned). Run your own slop check before showing me: if any sentence could appear in another agent's post, rewrite it first. Deliver as a Monday–Thursday schedule with a one-line note on why each post fits its day.
I hold this opinion about my market: [STATE YOUR ACTUAL TAKE]. Play devil's advocate first — give me the three strongest counterarguments a smart colleague would make. Let me respond to each out loud [I'LL TYPE MY RESPONSES]. Then write the opinion post from MY responses — my position, my reasoning, my words, sharpened by the pushback. Rules: never soften my take into "there are two sides" mush, never add opinions I didn't state, acknowledge the strongest counterpoint in one honest line because that's what confident people do. My voice, under 150 words.
Take this week's strongest post [PASTE] and translate it natively for: Facebook (conversational, room to breathe), Instagram (tight caption, hook first line, line breaks for air), LinkedIn (the professional-lesson angle, but still ME — no corporate voice creep), and a 30-second talking-head video script in my spoken rhythm with my actual phrases. The voice stays identical across all four — only the format changes. Flag which platform this particular story will hit hardest on and why.
Design my signature recurring series — the thing people in [MY AREA] associate with MY name. Based on my Voice Card, my four pillars, and what I most enjoy talking about [TELL IT], propose three series concepts with names in my voice (not "Market Monday" — something only I would call it). For each: the format, the weekly source material it runs on (ideally my Monday Memo), an example first post, and why it's defensible — meaning another agent copying it would obviously look like they copied me. I'll pick one; then write its first three installments.
Monthly review. Here are my posts from the last 30 days with their results [PASTE POSTS + rough numbers: reach, comments, DMs, and — most important — any real-world mentions: "someone brought up my post at a showing"]. Analyze what RESONATED, not just what performed: which pillar earned the most genuine replies, which post sounded most like me AND hit hardest, what the DMs and in-person mentions have in common. Attention that leads to conversations beats reach that doesn't. Give me: my top resonance pattern, one pillar to double down on, one experiment for next month, and the post from this month I should turn into a series.
From "any agent could have posted this" to unmistakable — in four weeks
Attention compounds — here's where it lands in the P&L
Signed today. Voice Card this week; the engine runs within thirty days.
Everyone has the same AI now.
Nobody else has my voice.
The feeds in my market are filling with content that is fluent, polished, and instantly forgettable — a thousand agents typing the same prompt and publishing the same average. I understand the temptation. Generic is fast, generic is safe, and generic is invisible.
I choose the other road. I taught the machine to sound like me — from my real emails, my real texts, the way I actually talk when a deal is on the line. I filter everything it drafts until it passes one test: would I say this, out loud, to a person I respect? And I feed it the one thing no model can generate — what actually happened to me this week. My Tuesday. My market. My take.
The tools stopped being the edge the day everyone got them. What's left is what was always the edge: a recognizable human being, showing up consistently, saying true things in their own words. AI didn't kill that game. It cleared the field for whoever still plays it.
Printed. Filtered. Injected. Published. Repeated.
The game of attention goes to the voice they remember.
"Because Sounding Like Everyone Else is NOT an Option!"
— Edmund Bogen · Edmund's Mastermind · Bogen.ai