Day 2: CMA Reports With AI in 30 Minutes

Goal

Build an end-to-end CMA workflow that pulls comps from your MLS, runs adjustments through AI with explicit logic you can defend at the listing appointment, and outputs a seller-ready narrative — in thirty minutes, every time.

What You'll Learn

How to structure your MLS comp export so AI can read it cleanly.
A prompt that runs square-footage, condition, location, and time adjustments with transparent reasoning.
How to generate the seller-facing narrative — high range, low range, recommended list price, days-on-market expectation.
The Fair Housing line in CMAs that most agents do not realize they are crossing.
A slide structure you can paste into Canva or PowerPoint in five minutes.

Yesterday we wrote listing copy. Today we work upstream — to the listing appointment itself. The CMA is the document that wins the listing, and most agents either over-engineer it (four hours, eighty slides, glazed seller eyes) or under-engineer it (three Zillow screenshots and a prayer). AI lets you split the difference: a thirty-minute, defensible, beautiful CMA every time.

The Comp Export and the AI Adjustment Pass

Step one is the data. Pull six to ten comps from your MLS — sold within the last six months, within a quarter mile if urban or a mile if suburban, within ten percent of the subject's square footage. Export to CSV. Open the CSV. Make sure the columns include address, sold price, sold date, beds, baths, square feet, lot size, year built, condition rating, and any unique features.

Now paste the data into your AI tool with this prompt:

You are a senior listing agent preparing a CMA.

SUBJECT PROPERTY
Address: [123 Maple Ave]
Beds/Baths: [4 / 2.5]
SqFt: [2,140]
Lot: [0.32 ac]
Year built: [1998]
Condition: [Good — updated kitchen 2023, roof 2024]
Special features: [finished basement, 3-car garage]

COMPS (paste CSV rows here)
[address, sold price, sold date, beds, baths, sqft,
lot, year, condition notes]
1) ...
2) ...

INSTRUCTIONS
1. For each comp, list adjustments line by line:
   - SqFt diff x $80/sqft (state your $/sqft logic)
   - Condition (excellent +5%, good 0%, fair -5%)
   - Lot size (note if material)
   - Time of sale (apply +/- 0.4% per month
     since sold)
   - Unique features (basement, garage stalls)
2. Show subtotal adjusted price per comp.
3. Drop the highest and lowest adjusted comps.
4. Give me: average of remaining, median,
   high range, low range, and recommended list.
5. Explain your $/sqft and time-adjustment
   assumptions in two sentences each.

The model will produce a clean adjustment table you can sanity-check in two minutes. The key is that you set the adjustment rules — the model does the arithmetic. That keeps you defensible: every number traces to a rule you chose.

Tip. Always tell the model to drop the highest and lowest adjusted comp. It removes outliers without you cherry-picking, and it gives you a cleaner story for the seller.

The Seller Narrative

Numbers don't win listings. Narrative does. After the adjustment pass, run a second prompt:

Using the CMA above, write the listing-appointment
narrative I will read to the seller. 250-320 words.

Cover:
- What the data is telling us right now
- Where the subject sits relative to the market
  (premium / at-market / value-priced)
- The case for the recommended list price
- The expected days-on-market range
- The two biggest risks (one market risk, one
  property-specific) and how we mitigate them
- A confident, plain-English close

Voice: experienced, direct, no jargon, no
"in this market" filler.

The output is the script you bring to the kitchen table. Read it once. Edit two sentences so it sounds like you. Done.

Fair Housing Act & NAR Article 10 — Pricing-Side Pitfalls

Where pricing meets civil rights. The Fair Housing Act (42 U.S.C. §3604) and the Equal Credit Opportunity Act extend beyond advertising into appraisal and pricing practices. NAR Code of Ethics Article 10, Standard of Practice 10-1, prohibits REALTORS® from "volunteering information regarding the racial, religious or ethnic composition of any neighborhood." If your CMA narrative says "this neighborhood is changing demographically and that's why values are rising," you have just violated SOP 10-1 and likely created Fair Housing exposure. AI will write that sentence for you if you let it. You cannot let it.

Specific things to scrub from any AI-generated CMA narrative:

Demographic commentary on neighborhoods. "Young professionals are moving in," "this area is becoming more diverse," "the school district is changing." Replace with property-level and market-level facts: median price trend, days-on-market trend, inventory months.
School-quality language tied to outcomes. Naming a school is fine; "a top-rated school district that protects values" risks pretextual discrimination claims. Cite the school name and a public rating source; let the seller draw the inference.
Steering language. "Buyers like you tend to prefer the north side" is steering — a separate Fair Housing violation under 24 CFR §100.70. CMAs are about the subject property's value, not where the seller's next buyer should live.
Algorithmic-bias issues. If you use AVMs (Zestimate, Redfin Estimate, RPR) as comps, note them as one input — not the anchor. HUD has begun investigating algorithmic appraisal bias; documenting that you reasoned beyond the AVM protects you.

Add this guardrail to the bottom of every CMA prompt you run from now on:

Do not include demographic, racial, religious,
familial, or national-origin commentary about the
neighborhood, schools, or expected buyers.
Property and market metrics only. Comply with
42 U.S.C. §3604 and NAR Code of Ethics Article 10
Standard of Practice 10-1.

Disclaimer. CMAs are not appraisals. State licensing law dictates what licensees may and may not do in pricing opinions — and several states (notably under USPAP-influenced rules) require specific disclosures. Confirm with your broker and state real-estate commission before delivering AI-assisted CMAs to clients.

Homework

Pull a real comp set from your MLS for a property you might list this quarter.
Run the adjustment prompt. Time yourself from CSV-ready to recommended list. Beat thirty minutes.
Run the narrative prompt. Read it aloud — does it sound like you?
Run the Fair Housing scrub on the narrative output. Note anything flagged.
Save the comp CSV, the adjustment output, and the narrative as a single PDF. That is your CMA template going forward.

← Day 1: Listings Course index Day 3: Lead Nurture →

CMA Reports With AI in 30 Minutes

Goal

What You'll Learn

The Comp Export and the AI Adjustment Pass

The Seller Narrative

Fair Housing Act & NAR Article 10 — Pricing-Side Pitfalls

Homework

Related courses