Skip to content
GuidesReal EstateMarketingFamilyLive EventseSports
All free classes

Free class · How to Stop AI Writing From Sounding Robotic

How do I make an AI draft sound less robotic before I send it?

Stop describing your tone and give the AI three to five messages you actually sent, the words you never use and your Fair Housing limits. Paste that Voice Kit with the draft, ask for a rewrite plus a list of facts to check, and read it yourself before it goes out.

Control room · 13 slides

Play the deck, one slide at a time.

The slides from the class. Use the buttons or your arrow keys, or press F for full screen. Every slide is also below as text.

Voice Kit Lab. AI Acceleration

Voice Kit Lab

Slide 1 of 13: Voice Kit Lab

Read all 13 slides as text

  1. 01 Voice Kit Lab

    AI Acceleration

    Real Estate

    INNOVATION LAB · NORTH · AUGUST 28, 2026

    Make AI sound like you. Not a listing robot.

    Presented by Build Things That Build Things

  2. 02 Build your Voice Kit. Rewrite three pieces. Stop sounding robotic.

    Minutes 0–25: Build Voice Kit Part 1 (voice samples from emails/texts you actually sent).

    Minutes 25–55: Build Voice Kit Parts 2–3 (do-not-say list + Fair Housing constraints).

    Minutes 55–85: Rewrite loop — take 3 pieces of your real work and rewrite them in your voice, live.

    Minutes 85–90: Before/after showcase. Homework: use it 5x this week.

  3. 03 Why AI sounds generic by default.

    You gave it a task. It needs your voice.

    Low perplexity: AI chooses "obvious" words. Humans use unpredictable, distinctive phrasing.

    Repetitive structures: AI follows formulaic patterns. Your writing has rhythm variation.

    Missing emotional nuance: AI is neutral. Your voice has warmth, humor, or directness baked in.

    No do-not-say filter: AI doesn't know you hate "stunning" or "motivated seller."

    Source: UCLA HumTech AI Detection Study; DivEye rhythmic unpredictability research.

  4. 04 Would your client recognize this as yours if your name wasn't on it?

    That is the only metric that matters.

    Pass: Client reads it and says "this sounds like you" or doesn't notice AI was involved.

    Fail: Client reads it and thinks you copy-pasted from the internet or another agent wrote it.

    The gap: Every generic AI output is a missed opportunity to reinforce your brand.

    The fix: Voice samples + do-not-say list = outputs that pass the client test.

  5. 05 Draft ≠ Send

    Every AI output requires human review before it leaves your hands.

    AI drafts. You verify facts, check Fair Housing compliance, review for client-specific constraints.

    The discipline: Treat every output as a draft. Speed without review creates compliance disasters.

    What you check: (1) Sounds like me, (2) Facts accurate, (3) Fair Housing clean, (4) Client-appropriate.

    The gate: If you wouldn't sign your name to it after reading it once, don't send it.

  6. 06 Three sections: Samples, Do-Not-Say, Constraints.

    Show your voice. Define your boundaries. Enforce compliance.

    Part 1: Voice Samples — 3-5 emails/texts you actually sent. AI learns your rhythm, word choices, openings/closings.

    Part 2: Do-Not-Say List — Words/phrases you refuse to use. Realtor clichés. Tone you avoid. Your brand boundaries.

    Part 3: Fair Housing Constraints — The non-negotiables. What you will never say. Compliance as voice, not rules.

    Usage: Paste all three parts + your draft. AI rewrites in your voice, respecting your boundaries.

  7. 07 Find 3 emails or texts you sent that sound like you.

    AI learns from what you show, not what you describe.

    What to pull: Emails or texts to clients, other agents, vendors — anything you sent that felt authentic.

    What to look for: Messages people complimented. Follow-ups that got replies. Posts clients engaged with.

    Length: 3–8 sentences per sample. Full emails, not snippets. AI needs rhythm, not keywords.

    Right now: Open your sent folder. Find 3. Paste them into a new text file labeled "Voice Kit."

  8. 08 Trade Voice Kits with the person next to you.

    What do you notice about their voice? Tell them.

    Instructions: Swap files. Read their 3 samples. Notice: sentence length, word choices, formality, humor, distinctive phrases.

    Tell them: "I notice you [open with a question / use short punchy sentences / avoid jargon / sound warm]."

    What this does: You can't always see your own patterns. Another agent can. Use their observations.

    Time limit: 3 minutes. One observation minimum. Specifics, not "sounds good."

  9. 09 Words and phrases you refuse to use.

    Your brand boundaries. The clichés you avoid.

    Realtor clichés: "Stunning," "motivated seller," "don't miss out," "dream home" — overused or legally risky.

    Tone you avoid: Corporate jargon, fake enthusiasm, overly formal legalese, whatever feels inauthentic to you.

    Your list will be unique: One agent avoids humor. Another leans into it. There is no universal do-not-say.

    Right now: Add a section to your Voice Kit file. Title it "Do-Not-Say List." Brainstorm 5–10 items.

  10. 10 How to use your Voice Kit every time.

    Paste Kit + draft. AI rewrites in your voice.

    Step 1: Paste your entire Voice Kit (samples + do-not-say + constraints) into ChatGPT or Claude.

    Step 2: Paste the draft you want rewritten (generic AI output, or an old email you want improved).

    Step 3: Use this prompt: "Rewrite the draft below in my voice using the samples and constraints above. Flag any facts you cannot verify and any Fair Housing concerns."

    Step 4: Review output. Does it sound like you? Facts accurate? Fair Housing clean? Edit as needed.

  11. 11 Pick 3 pieces from your real work. Rewrite them in your voice.

    The AI draft you hate. An email you want improved. A listing description. Your choice.

    What to bring: The AI draft you hate, or 2–3 emails/listings you want to improve. Your actual work.

    The loop: Paste Voice Kit → paste draft → run Style Match prompt → review → does it sound like you?

    If no: Diagnose which section of your Voice Kit is weak. Revise the Kit, not just the output.

    Time: 20 minutes. Rewrite 3 pieces. Walk out with usable output or clear diagnosis of what to fix.

  12. 12 Volunteers: share your original and your rewrite.

    Group votes: would you send this? What still needs human touch?

    Instructions: Read your original AI output or email. Then read the Voice Kit rewrite. Group: would you send this?

    What we're calibrating: The "sounds like me" test. The draft-versus-send discipline. What still needs editing.

    Honesty required: If it still sounds generic, say so. If it's sendable, say that. No politeness votes.

    Key insight: Even voice-matched output needs a final fact-check and client-specific review. Speed, not autopilot.

  13. 13 Voice Kit → Style Match → Rewrites.

    Three parts. One prompt. Reusable forever.

    Voice Kit (3 sections): Samples of your actual writing + do-not-say list + Fair Housing constraints.

    Style Match prompt: Paste Kit + draft. AI rewrites in your voice, flags facts to verify, checks Fair Housing.

    3 finished rewrites: From your real work. Ready to use (after human review).

    The reflex: Bad output? Diagnose which Kit section is weak. Fix the Kit, not the prompt.

Class replay · not verified

No verified replay is available.

Voice Kit Lab was taught live at Innovation Lab North, Compass Lakeside, Hendersonville on August 28, 2026. We have not located a verified recording of this class.

The slides above, the steps below and the workbook carry the lesson. The workbook's prompts and its Voice Kit template are printed here in full.

Each class was taught live by Ryan Wanner to working real estate agents.

Playback · before the fix

Why a draft comes back flat.

You gave the AI a task. You didn't give it your voice. With nothing else to go on, it does four things.

Take 1 The AI draft

Same weight on every word. Same sentence shape, start to finish.

Take 2 The rewrite in your voice

Short lines and long ones. Your words, your openings, your sign-off.

A drawing of rhythm, not a measurement.

  1. It picks the obvious word.

    AI leans toward the most expected next word. People don't. Your turns of phrase are what make your writing yours.

  2. It repeats the same shapes.

    Generic output falls into a formula. Your writing has rhythm. A short sentence. Then a longer one that takes its time.

  3. It stays neutral.

    Unless you show it otherwise, it has no warmth, no dry humor and no bluntness. Your voice has at least one of those.

  4. It has no do-not-say filter.

    It doesn't know you'd never write “stunning” or “motivated seller.” It will use every cliché in the business until you name them.

These four come from the class deck. Slide 3 · Why AI sounds generic by default.

The playback test

Would your client recognize this as yours if your name wasn't on it?

Pass
They say it sounds like you, or never notice AI helped.
Fail
They think you pasted it from the internet, or another agent wrote it.

The console · 8 channels

Eight steps from a flat take to your voice.

Work the channels in order. Channels 1 to 4 build your Kit once. Channels 5 to 7 run on every draft.

  1. Step 1: Identify what makes the draft feel generic

    Read the draft aloud and note repeated phrasing, flat sentence rhythm or clichés. Compare those passages with your permitted voice samples. Missing voice context is one possible cause; the task, source material or model may also need attention.

  2. Step 2: Choose three to five permitted voice samples

    Pick examples that sound like you and are safe to use in the approved account. Use fictionalized or minimized passages that preserve useful rhythm without private client details. Save the reviewed samples in your Voice Kit; do not paste whole client messages merely because you wrote them.

  3. Step 3: List the words you never use

    Under your samples, add a do-not-say list of 5 to 10 words, phrases and tones you refuse to use. Start with the clichés in the robotic draft. Your list is yours. One agent avoids humor, and another leans into it.

  4. Step 4: Write your Fair Housing limits into the Kit

    Add a third section. Describe the property and its features, never who it's for or who lives nearby. No protected-class language, even when it sounds kind. Fit it to your state and your brokerage's policy.

  5. Step 5: Load the reviewed Kit and one permitted draft

    Use an approved account that supports the selected inputs. Supply the current Voice Kit and a fictional or permitted draft, then request a rewrite with no invented facts, facts to verify and possible language issues. Check the actual output against the sources; the model's self-review is not proof.

  6. Step 6: Diagnose one mismatch and test a repair

    Compare the disappointing passage with the supplied samples, facts and instructions. Change the section responsible, then rerun and compare the actual result. Do not assume the Kit is always the cause or that another run must improve it.

  7. Step 7: Run four checks before authorized use

    Compare voice with samples, verify factual claims, review possible Fair Housing issues under brokerage policy, and check suitability for the intended recipient. AI reporting no issue is not legal clearance. Resolve uncertainties before any authorized send.

  8. Step 8: Test the Kit on a few relevant drafts

    Try it on fictional or permitted examples of the work you actually do. Record the changes needed, update one useful rule and compare the next result. Continued use alone does not improve the Kit; retain changes only when reviewed results support them.

Use the class workbook · Get the supporting files

Read and work through the lesson here. Downloadable files use your free student account.

The limiter

Fair Housing limits, set before the first take.

Fair Housing isn't a check you run at the end. It's part of how you write. Put these limits in the third section of your Kit, so the AI works inside them from the first draft.

Never characterize

  • Who a property or neighborhood is “perfect for”
  • A neighborhood by who lives there
  • Schools as good or bad
  • Anything tied to a protected class, even when it sounds kind

Describe features instead

  • “Parks and sidewalks,” not “family-friendly”
  • “Near transit and dining,” not “young professional area”
  • Amenities, not residents
  • Let buyers decide for themselves what fits

The steering test: if a sentence suggests who should or shouldn't live somewhere, rewrite it or delete it. Fit these limits to your state and your brokerage's policy. The limits make risky words less likely to show up. They don't clear a draft. You're still responsible for every word you send.

Cue marks · 8 lines from the slides

Lines worth a second look.

Every line here is copied word for word from a class slide. No verified replay has been located, so the slides are the record. Each cue opens its slide on the monitor above.

  1. Slide 3 · Why AI sounds generic by default.
    “You gave it a task. It needs your voice.”

    A robotic draft usually means the AI got the job and nothing about how you write.

  2. Slide 3 · Why AI sounds generic by default.
    “AI doesn't know you hate "stunning" or "motivated seller."”

    Clichés show up because nobody told the AI which words you never use.

  3. Slide 4 · Would your client recognize this as yours if your name wasn't on it?
    “Would your client recognize this as yours if your name wasn't on it?”

    The test for a rewrite is whether your client would know it was yours without your name on it.

  4. Slide 7 · Find 3 emails or texts you sent that sound like you.
    “AI learns from what you show, not what you describe.”

    Messages you really sent fix a robotic draft better than tone words like friendly or professional.

  5. Slide 7 · Find 3 emails or texts you sent that sound like you.
    “3–8 sentences per sample. Full emails, not snippets. AI needs rhythm, not keywords.”

    Whole messages carry your rhythm and how you open and close, which short snippets lose.

  6. Slide 10 · How to use your Voice Kit every time.
    “Rewrite the draft below in my voice using the samples and constraints above. Flag any facts you cannot verify and any Fair Housing concerns.”

    The Style Match rewrite asks for your voice and, in the same request, a list of facts and Fair Housing concerns for you to check.

  7. Slide 11 · Pick 3 pieces from your real work. Rewrite them in your voice.
    “Diagnose which section of your Voice Kit is weak. Revise the Kit, not just the output.”

    When a rewrite still misses, the fix goes into the Kit, so the next draft starts closer to you.

  8. Slide 12 · Volunteers: share your original and your rewrite.
    “Even voice-matched output needs a final fact-check and client-specific review. Speed, not autopilot.”

    A rewrite that sounds like you still needs your read for facts and fit before it goes out.

Clipping · 6 faults

Six ways the mix goes wrong.

  1. Describing your tone instead of showing it

    “Friendly and professional” fits every agent, so it gives the AI nothing to copy. Messages you really sent show it your rhythm, how you open and close, and how formal you are.

  2. Rewriting the prompt instead of the Kit

    A new prompt throws away what the last draft taught you. Find the weak part of your Kit, fix that one thing and run it again. The next draft starts from the fix.

  3. Pasting snippets as voice samples

    Subject lines and one-liners don't show rhythm. Paste full messages of 3 to 8 sentences so the AI can see how you open, build and close.

  4. Checking Fair Housing only at the end

    If your limits aren't in the Kit, a rewrite can still say who a home is “perfect for.” Put them in from the start, and still read every draft for steering yourself.

  5. Sending it because it sounds like you

    A rewrite in your voice can still get a date, a price or a feature wrong. Check every item on the facts list against the source before anything goes out.

  6. Buying a tone menu to fix it

    A writing tool's tone library is someone else's guess at how agents sound. A plain text Voice Kit you own works in ChatGPT or Claude, and it goes with you if you switch.

Up next in the session

Where to take your voice next.

All free classes

On the desk · the workbook

The whole workbook, on the desk.

What you build A reusable Voice Kit (writing samples + do-not-say list + Fair Housing constraints) and Style Match discipline for AI-generated content that sounds authentically you.

Pick your room.

The class works in either one. Use the one you already have.

ChatGPT (chatgpt.com)

OpenAI's foundation model. Check the current plan, file support and usage limits for your account; practice can begin with manual review. Paste your Voice Kit, paste a draft, get a rewrite in your voice.

Best for
Style Match rewrites, listing copy, client emails, social posts. Works with voice samples only — no vendor tools required.
Cost
Check the current plan, file support and usage limits for your account; practice can begin with manual review.
  1. Go to chatgpt.com and sign up or sign inCheck the current plan, file support and usage limits for your account; practice can begin with manual review.
  2. Start a new chatEach Voice Kit rewrite should start in a fresh chat so the model has clean context.
  3. Paste your Voice Kit at the top of every Style Match promptVoice Kit = voice samples + do-not-say list + Fair Housing constraints. Paste it before the draft you want rewritten.

The first five minutes

  1. Open a new chat and paste one email or text you sent that sounds like you. Ask ChatGPT to describe your voice.
  2. Paste the Style Match prompt from the class prompt pack.
  3. Paste a real draft (listing, email, social post) and run the Style Match workflow end to end.
  4. Review the Four-Gate Test: Sounds like me, Facts accurate, Fair Housing clean, Sendable.
  5. Keep every output in draft. A draft is not a send.

Claude (claude.ai)

Anthropic's foundation model. Check the current plan, file support and usage limits for your account; practice can begin with manual review. Paste your Voice Kit, paste a draft, get a rewrite with reasoning.

Best for
Style Match rewrites, complex Fair Housing checks, nuanced voice analysis. Works with voice samples only — no vendor tools required.
Cost
Check the current plan, file support and usage limits for your account; practice can begin with manual review.
  1. Go to claude.ai and sign up or sign inCheck the current plan, file support and usage limits for your account; practice can begin with manual review.
  2. Start a new conversationEach Voice Kit rewrite should start in a fresh conversation so the model has clean context.
  3. Paste your Voice Kit at the top of every Style Match promptVoice Kit = voice samples + do-not-say list + Fair Housing constraints. Paste it before the draft you want rewritten.

The first five minutes

  1. Open a new conversation and paste one email or text you sent that sounds like you. Ask Claude to describe your voice.
  2. Paste the Style Match prompt from the class prompt pack.
  3. Paste a real draft (listing, email, social post) and run the Style Match workflow end to end.
  4. Review the Four-Gate Test: Sounds like me, Facts accurate, Fair Housing clean, Sendable.
  5. Keep every output in draft. A draft is not a send.

The channel strip: your Voice Kit.

One plain text file, three sections. Show your voice, set your limits, and write Fair Housing in from the start.

  1. In

    Voice samples

    Paste 3 to 5 emails or texts you actually sent that sound like you. Full messages of 3 to 8 sentences, not snippets. The AI picks up your rhythm, openings, closings and word choices from real examples.

  2. Gate

    Do-not-say list

    Name every word, phrase and tone you refuse to use. Real estate clichés, corporate jargon, fake enthusiasm. Start with 5 to 10 items and add one every time a draft makes you wince.

  3. Limit

    Fair Housing limits

    Write the non-negotiables into the Kit itself. Describe the property, never who it's for. Describe amenities, never the residents. No protected-class language, even when it sounds positive.

  4. Out

    Use it every time

    Paste all three parts, then the draft you want rewritten. The same Kit works for listing copy, follow-up emails, texts and social posts. Build it once and keep tuning it.

Pulling your samples

Open your sent folder and your texts. Look for the note to a client after a showing, the text to a seller about offer timing, the follow-up that got a reply, or the post people liked. Paste them in whole. The AI needs your openings and closings, not a keyword list. Take out client names, numbers and private details first.

Then hand the file to a colleague and ask what they notice. Do you open with a question? Keep sentences short? Skip the jargon? You can't always see your own patterns. Someone else can.

voice-kit-template.md
# Voice Kit

## Voice Samples (3-5 examples)
Paste 3-5 emails or texts you actually sent that sound like you. Full messages, not snippets. AI learns rhythm from real examples.

**Example 1:**
[Paste here]

**Example 2:**
[Paste here]

**Example 3:**
[Paste here]

---

## Do-Not-Say List
Every word or phrase you refuse to use. Ask AI to avoid these, then check the output for every listed phrase.

- [e.g., stunning, motivated seller, don't miss out]
- [Add your banned phrases]

---

## Fair Housing Constraints
Language constraints to check on every draft. They do not guarantee compliance; follow your brokerage review process.

**Never characterize:**
- Who a property is "perfect for"
- Neighborhood demographics or who lives there
- Schools or school quality, including ratings

**Never mention protected classes** (even positively): race, color, religion, sex, national origin, familial status and disability, plus any classes your state adds, such as age.

**The Steering test:** If a sentence suggests who should or shouldn't live somewhere, rewrite or delete it.

The remix: four prompts.

Style Match is the remix. Paste your whole Kit, then one draft. If the remix still misses, don't write a new prompt. Find the weak channel and fix that one thing.

  • Still sounds generic

    Fix the input (voice samples)

    Your samples are too short or too alike. Add a fuller message that shows the warmth, humor or directness that's missing.

  • A cliché slipped in

    Fix the gate (do-not-say list)

    It isn't on your do-not-say list yet. Add the exact phrase. Your list grows every week you use it.

  • Risky language showed up

    Fix the limiter (Fair Housing limits)

    Your Fair Housing section is too vague. Name the pattern the draft used and the feature-based way to say it instead.

Style Match: rewrite any draft in your voice

Patch inVoice Kit (3 sections) + the draft you want rewritten

Prompt
I am a real estate agent. Below is my Voice Kit (voice samples + do-not-say list + Fair Housing constraints), followed by a draft I want rewritten in my voice.

[PASTE YOUR VOICE KIT HERE]

---

**Draft to rewrite:**
[PASTE DRAFT HERE]

---

Rewrite the draft in my voice, using the rhythm and phrasing patterns from my voice samples. Respect every item on my do-not-say list. Apply my Fair Housing constraints throughout. Do not add facts I did not provide.

Respond in exactly these blocks:

REWRITTEN VERSION
The full rewrite in my voice. No facts added, no do-not-say violations, Fair Housing clean.

FACTS TO VERIFY
Every factual claim I should verify before sending: dates, prices, features, timelines, anything that could be wrong.

FAIR HOUSING CHECK
Any sentence that might characterize who a property is for, mention protected classes, or suggest who should or shouldn't live somewhere. If you detect no issue, say NO ISSUE DETECTED. This is a limited language check, not legal clearance.

WHAT I MATCHED FROM YOUR VOICE
The specific patterns from my voice samples you used: sentence rhythm, opening style, closing style, distinctive phrasing.

A draft is not a send. Review every fact before this leaves your hands.

What comes back

  • REWRITTEN VERSION / FACTS TO VERIFY / FAIR HOUSING CHECK / WHAT I MATCHED FROM YOUR VOICE.
  • The rewrite uses patterns from voice samples, not generic real estate voice.
  • Every factual claim lands in FACTS TO VERIFY — no skips.
  • Fair Housing check flags possible risky language or reports NO ISSUE DETECTED. Neither result is legal clearance; review the actual text under brokerage policy.
  • Compare the claimed voice patterns with exact passages in the supplied samples and the rewrite. The model's explanation alone does not prove that it read or used the samples.

Paste your Voice Kit once at the top, then paste the draft below. One draft per prompt. Run this in a fresh chat so the model has clean context.

Diagnose: which Kit section is weak

Patch inVoice Kit + the disappointing rewrite

Prompt
I ran Style Match and the output disappointed me. Below is my Voice Kit and the rewrite AI produced.

[PASTE YOUR VOICE KIT HERE]

---

**AI rewrite:**
[PASTE REWRITE HERE]

---

Diagnose which Voice Kit section is weak. Respond in exactly these blocks:

VOICE SAMPLES DIAGNOSIS
Do my samples show enough rhythm variation, distinctive phrasing, and emotional tone? If they are too short or too similar, tell me.

DO-NOT-SAY DIAGNOSIS
Did the rewrite violate my do-not-say list? If so, which phrases appeared and why did AI miss them?

FAIR HOUSING DIAGNOSIS
Did the rewrite characterize who a property is for, mention protected classes, or suggest who should or shouldn't live somewhere? If so, quote the problem.

RECOMMENDED FIX
Which Kit section needs work, and what specific addition or change would strengthen it?

What comes back

  • VOICE SAMPLES DIAGNOSIS / DO-NOT-SAY DIAGNOSIS / FAIR HOUSING DIAGNOSIS / RECOMMENDED FIX.
  • The fix targets one Kit section with a specific addition or change.
  • Diagnosis identifies the weak section, not vague advice to 'be more specific.'

Run this when output disappoints. Fix the Kit, not the prompt. One diagnosis, one targeted fix.

Expand voice samples: find more of your voice

Patch inYour sent folder or text messages

Prompt
I am a real estate agent building my Voice Kit. I have 3 voice samples but I want to expand them. Below are 5-10 more emails or texts I sent.

[PASTE ADDITIONAL SAMPLES HERE]

---

Analyze these samples and respond in exactly these blocks:

WHICH SAMPLES TO KEEP
The 3-5 samples that show the clearest rhythm, distinctive phrasing, and emotional range. Explain why each one earns a spot.

WHAT YOUR VOICE DOES
The patterns I should see across all my samples: sentence rhythm, opening/closing style, humor or warmth or directness, distinctive vocabulary.

WHAT IS STILL MISSING
If my samples lack variation (all short emails, all formal tone, all one type of message), tell me what to add next time.

What comes back

  • WHICH SAMPLES TO KEEP / WHAT YOUR VOICE DOES / WHAT IS STILL MISSING.
  • Keeps 3-5 samples, not all of them.
  • Voice patterns are specific: sentence rhythm, opening style, distinctive phrasing.
  • Missing element is named so agent knows what to pull next.

Run this when you have too many samples or want to confirm which ones matter. Keeps the strongest 3-5.

Fair Housing check: scan any draft for risky language

Patch inVoice Kit Fair Housing constraints + the draft

Prompt
I am a real estate agent. Scan the draft below for Fair Housing violations.

**Draft:**
[PASTE DRAFT HERE]

---

Respond in exactly these blocks:

NO ISSUE DETECTED OR FLAGGED
If you detect no issue in this language check, say NO ISSUE DETECTED. If risky language appears, say FLAGGED. Neither result guarantees compliance; human and brokerage review still apply.

RISKY LANGUAGE (if any)
Quote every sentence that characterizes who a property is for, mentions protected classes, or suggests who should or shouldn't live somewhere.

SAFE ALTERNATIVE
For each flagged sentence, suggest a neutral, property-focused alternative for human review. Do not claim the alternative is legally compliant.

THE STEERING TEST
If any sentence fails the Steering test (does this suggest who should or shouldn't live here?), explain why and rewrite it.

What comes back

  • NO ISSUE DETECTED OR FLAGGED / RISKY LANGUAGE (if any) / SAFE ALTERNATIVE / THE STEERING TEST.
  • Flagged sentences are quoted exactly.
  • Review proposed alternatives for the underlying issue; softer wording alone does not resolve it. Escalate unresolved questions through the brokerage review process.
  • Steering test explains the why, not just the fix.

Use this as one language review aid before sending. It can miss issues; inspect the actual draft and escalate unresolved questions.

Before anything leaves the room.

Five checks from the workbook. Every light on, or it stays a draft.

  • Draft ≠ Send. Every AI output requires human review before it leaves my hands.

    I verify facts, check Fair Housing compliance, and confirm client-appropriateness. Check the actual wording and resolve uncertain claims through the brokerage review process before sending.

  • I never let AI characterize who a property or neighborhood is suited for.

    No 'perfect for families,' 'great for retirees,' 'ideal for young professionals.' AI describes the property, not its audience.

  • I never let AI mention protected classes, even positively.

    No references to families, age, religion, national origin, disability, familial status, even in complimentary language.

  • I run the Four-Gate Test on every rewrite before sending.

    1. Sounds like me. 2. Facts accurate. 3. Fair Housing clean. 4. Sendable. If it fails any gate, I diagnose which Kit section needs work.

  • I verify every factual claim AI generates before anything leaves draft.

    Dates, prices, features, timelines, property facts. AI invents plausible-sounding details. I check every one.

Your week of sessions.

  1. TODAY

    Build your Voice Kit: 3 voice samples + do-not-say list + Fair Housing constraints

    Pull 3 emails or texts you sent that sound like you. List every cliché you forbid. Add the Fair Housing constraints from class. Save as one file.

  2. TOMORROW

    Run Style Match on one real draft

    Paste your Voice Kit, paste a draft (listing, email, social post), and run the Style Match prompt. Run the Four-Gate Test before sending.

  3. DAY 3

    Diagnose one disappointing rewrite

    When output disappoints, run the diagnosis prompt. Fix the weak Kit section, rerun Style Match and compare the actual result.

  4. DAY 5

    Expand your voice samples

    Pull 5-10 more emails or texts you sent. Run the sample expander prompt to identify the strongest 3-5. Replace weak samples in your Kit.

  5. ONGOING

    Draft ≠ Send. Run the Four-Gate Test on every AI rewrite before it leaves your hands.

    1. Sounds like me. 2. Facts accurate. 3. Fair Housing clean. 4. Sendable. If it fails any gate, fix it before sending.

Every prompt in the pack.

The class prompt pack, printed in full. Its Style Match prompt is a shorter take on the one above. An AI check can flag a risky line. It can't clear a draft, so the final read is yours.

Open all 10 prompts from the pack

  1. 1. The Style Match Prompt (Primary Workflow)

    Use this every time you want AI to write in your voice.

    Pack 01
    Rewrite the draft below in my voice using the samples and constraints above.

Match my:
- Sentence rhythm and length
- Word choices and vocabulary level
- Opening/closing style
- Formality/humor balance
- Any distinctive phrases I use

Follow my do-not-say list strictly. Apply my Fair Housing constraints. Flag any facts you cannot verify from the original. Flag any potential Fair Housing concerns even if I didn't catch them.

OUTPUT FORMAT:
Rewritten version:
[your rewrite]

Facts to verify:
[list any claims you cannot confirm]

Fair Housing check:
[any concerns or "Clean"]

ORIGINAL DRAFT:
[paste your draft here]

  2. 2. Voice Kit Diagnosis Prompt

    Use this when output still doesn't sound like you. Diagnose which section of your Kit is weak.

    Pack 02
    I'm building a Voice Kit so AI can write in my voice. I've provided:
- Voice samples (3-5 examples of my actual writing)
- Do-not-say list (words/phrases I avoid)
- Fair Housing constraints

The AI output I'm getting still doesn't sound like me. Analyze my Voice Kit below and tell me:
1. What patterns are missing from my voice samples?
2. What clichés/phrases does my writing use that I didn't forbid?
3. Are my Fair Housing constraints clear and specific enough?
4. What should I add or revise?

MY VOICE KIT:
[paste your entire Voice Kit]

EXAMPLE OUTPUT THAT DOESN'T SOUND LIKE ME:
[paste the AI output you don't like]

WHAT I EXPECTED:
[describe what you wanted instead]

  3. 3. Voice Sample Expansion Prompt

    Use this to generate more voice samples from existing writing when you need more examples.

    Pack 03
    Below are 3 samples of my actual writing. Analyze them and tell me:
1. Sentence structure patterns (length, rhythm, variety)
2. Distinctive word choices or vocabulary level
3. How I open and close messages
4. Tone (formal/casual, warm/direct, humor/serious balance)
5. Any recurring phrases or stylistic quirks

Then recommend: what additional types of writing samples should I add to make my Voice Kit more complete?

MY SAMPLES:
[paste your 3 voice samples]

  4. 4. Fair Housing Constraint Check

    Use this to flag possible language issues before your own and brokerage review. AI can miss issues; neither an unflagged draft nor a suggested rewrite is legal clearance.

    Pack 04
    Review the text below for Fair Housing compliance. Flag any language that:
- Characterizes who a property or neighborhood is "for"
- Describes neighborhoods through demographic proxies
- Uses protected-class language (race, color, religion, sex, national origin, familial status, disability)
- Could be interpreted as steering

If you find concerns, suggest neutral, feature-based alternatives.

TEXT TO REVIEW:
[paste your draft]

  5. 5. Do-Not-Say List Expansion

    Use this to find clichés you're using but haven't explicitly forbidden.

    Pack 05
    I'm building a do-not-say list of realtor clichés and phrases I want to avoid in my writing. Review the samples below and identify:
1. Common realtor clichés I'm currently using
2. Overused industry phrases
3. Words/phrases that sound generic or inauthentic
4. Tone patterns I should avoid

MY WRITING SAMPLES:
[paste 3-5 of your actual emails/listings]

CLICHÉS I ALREADY AVOID:
[paste your current do-not-say list]

  6. 6. Voice-Matched Email Templates

    Use this to create email templates in your voice for common scenarios.

    Pack 06
    Using my Voice Kit below, create email templates for these common scenarios:
1. Follow-up after showing (buyer)
2. Update to seller on market feedback
3. Congratulations on offer acceptance
4. Request for testimonial/review
5. [Your scenario]

Each template should:
- Match my voice exactly (sentence rhythm, word choices, openings/closings)
- Follow my do-not-say list
- Be Fair Housing compliant
- Include [BRACKETS] for variable details I fill in

MY VOICE KIT:
[paste your entire Voice Kit]

  7. 7. Listing Description Voice Match

    Use this to rewrite listing descriptions in your voice.

    Pack 07
    Rewrite the MLS listing description below in my voice.

Requirements:
- Match my sentence rhythm, word choices, and tone
- Follow my do-not-say list (no clichés I avoid)
- Apply my Fair Housing constraints (describe features, not ideal residents)
- Keep it within your MLS's character limit
- Flag any facts I need to verify with seller

MY VOICE KIT:
[paste your entire Voice Kit]

CURRENT LISTING DESCRIPTION:
[paste the draft you want improved]

  8. 8. Social Post Voice Match

    Use this to rewrite social media posts in your voice.

    Pack 08
    Rewrite the social post below in my voice for [Instagram/Facebook/LinkedIn].

Requirements:
- Match my voice (see samples below)
- Follow my do-not-say list
- Platform-appropriate length and format
- Include call-to-action if relevant
- Fair Housing compliant

MY VOICE KIT:
[paste your entire Voice Kit]

ORIGINAL POST:
[paste the draft you want improved]

PLATFORM:
[Instagram/Facebook/LinkedIn/TikTok]

  9. 9. Before/After Comparison

    Use this to evaluate whether a rewrite actually sounds like you.

    Pack 09
    Compare these two versions and tell me:
1. Which one sounds more like my voice samples?
2. What specific differences make one better than the other?
3. Does the rewritten version follow my do-not-say list?
4. Is the rewritten version Fair Housing compliant?

MY VOICE SAMPLES:
[paste 2-3 samples]

ORIGINAL:
[paste original draft]

REWRITTEN VERSION:
[paste AI rewrite]

  10. 10. Voice Kit Refinement Workflow

    Use this monthly to improve your Voice Kit based on what you've learned.

    Pack 10
    I've been using my Voice Kit for [timeframe]. Help me refine it.

What I've noticed:
- [List patterns in output you like/dislike]
- [List words/phrases AI keeps using that you don't like]
- [List any Fair Housing concerns you caught]

MY CURRENT VOICE KIT:
[paste your entire Voice Kit]

EXAMPLES OF RECENT OUTPUT:
[paste 2-3 recent AI rewrites]

Recommend:
1. Voice samples I should add or replace
2. Do-not-say items I should add
3. Fair Housing constraints I should clarify
4. Any patterns I'm missing

How the pack is meant to be used

  • Always paste your complete Voice Kit before using these prompts
  • Start with the Style Match prompt (#1) for daily use
  • Use diagnosis prompts (#2, #3, #5) when output doesn't sound right
  • Run Fair Housing check (#4) on anything client-facing before sending
  • Refine monthly using prompt #10

Stems · 5 files

Session files to take home.

Create a free student portal account to download this class's files.

  1. Voice Kit Template

    Structured template for capturing 3-5 voice samples, your do-not-say list, and Fair Housing constraints. Save as plain text or markdown.

    Markdown · 4 KB · Free account

    Get the file: Voice Kit Template

  2. Voice Kit Prompt Pack

    10 essential prompts: Style Match, diagnosis, sample expansion, Fair Housing check, Kit refinement, and voice sample evaluation.

    Markdown · 7 KB · Free account

    Get the file: Voice Kit Prompt Pack

  3. Before/After Rubric

    Four-Gate Test for every AI rewrite: Sounds like me, Facts accurate, Fair Housing clean, Sendable. Includes diagnosis trees and common failure patterns.

    Markdown · 7 KB · Free account

    Get the file: Before/After Rubric

  4. Fair Housing Voice Check

    Reference sheet: language to avoid (characterizing properties/neighborhoods, protected classes), safe alternatives, and the Steering test for every draft.

    Markdown · 8 KB · Free account

    Get the file: Fair Housing Voice Check

  5. Class skill card

    The class method on one page, written so you can hand it to your AI as instructions.

    Markdown · 15 KB · Free account

    Get the file: Class skill card

Questions about using this class

Practical answers from the lesson and workbook.

How many writing examples should I start with?

The class starts with three to five messages you actually wrote. Choose examples that represent the work you want to produce and remove private details. Review the resulting voice description rather than assuming it captured you accurately.

What belongs in my voice kit besides tone words?

Include examples of sentence rhythm, words you use, phrases you avoid, audience context and factual or compliance limits. Concrete examples are easier to check than a broad instruction such as 'sound authentic'.

What if the rewritten draft still sounds wrong?

Name the exact mismatch, correct the kit and run the same draft again. Compare the original, revision and examples side by side. Keep the facts stable while you work on the voice.

Does matching my voice make a real estate draft ready to publish?

No. Voice is one check. Verify the facts, relevance, permission and sensitive or discriminatory wording, then review the exact version before approving publication.

Your voice, every month

Your voice, on content you didn't have to write.

A Voice Kit fixes one draft at a time. Content Machine builds your voice into a context folder during the install, then makes at least 20 finished content assets a month in that voice, and nothing is published until you approve it.

Liner notes

Where the Fair Housing limits come from.