How Do Agents Build an AI Buyer Report in Minutes?
By Emily Terrell — Top Coach and Speaker at Tom Ferry International. 4 years coaching at Tom Ferry, 9 years prior as a client.
An AI buyer report is a buyer-side CMA: the same comps you’d use to price a listing, reframed to answer what this buyer should offer and why. With an AI CMA tool pulling comps and a saved prompt drafting the narrative, the build takes minutes. This guide covers the five-part structure, the prompt, and the compliance lines.
Key Takeaways
- A buyer report uses the same comps as a listing CMA but answers a different question: what to offer, not what to list at.
- AI handles comp scoring, price-per-square-foot math, and the plain-language narrative. You handle comp selection, offer advice, and the final check.
- An on-market Zestimate isn’t an independent check on list price, because Zillow counts the listing price among its inputs.
- In Texas, a buyer CMA carries the same TREC estimated-sale-price statement as a listing CMA.
- Neighborhood “color” about schools, safety, or who lives there is where AI-drafted buyer reports drift into steering. Leave it out.
What is an AI buyer report?
An AI buyer report is a one-to-two-page analysis built for a buyer before they write an offer. It covers four things:
- where the list price sits against recent comparable sales
- an estimated sale price range
- the seller-side signals visible in the data
- the questions to verify before signing anything
AI tools assemble the comps and draft the explanation. The agent curates, checks, and advises.
Why this matters for real estate agents
Your buyer is probably not competing on equal footing. According to NAR’s 2025 Profile of Home Buyers and Sellers (November 2025), the share of first-time home buyers dropped to a record-low 21%, and the typical first-time buyer is now 40 years old. The same report found 30% of repeat buyers paid all cash.
“Unfolding in the housing market is a tale of two cities.”
— Jessica Lautz, NAR Deputy Chief Economist
A first-time buyer putting 10% down against an all-cash repeat buyer can’t afford to guess. They need a number they can defend to themselves, to their lender, and to the listing agent.
Their first instinct is to check the Zestimate. Zillow reports a median error of 1.9% for homes on the market and 7.0% for homes that aren’t. That 1.9% looks reassuring until you read Zillow’s own list of inputs, which includes on-market data such as the listing price. The estimate your buyer is using to test the list price is partly built from the list price.
Here’s the thing nobody wants to tell you: that’s why a buyer report beats a screenshot.
How is a buyer CMA different from a listing CMA?
The data is the same. The question is not.
| Listing CMA | Buyer report | |
| Core question | What should this home list be? | What should my buyer offer? |
| Key signals | Competing actives, absorption | Days on market, price cuts, list-to-sale ratios |
| Output | Pricing strategy | Offer range and terms strategy |
If you’re building the seller side, the AVM and accuracy conversation is covered in Can AI Accurately Price a Home? This post stays on the buyer’s side.
“A listing CMA answers what a home should sell for. A buyer report answers what my client should pay and why the seller will say yes. Same comps, different question, and AI gets me from the comp set to that answer in minutes.”
— Emily Terrell, Tom Ferry Coach
The five-part AI buyer report framework
What does the AI need before it starts?
Clean inputs, and nothing it shouldn’t have. You need four things:
- the subject listing’s facts and price history
- your comp export
- the buyer’s practical constraints: maximum price, financing type, timeline, and must-have contingencies
Leave out the buyer’s name, income, and pre-approval letter. The analysis doesn’t need them. The account you’re typing into matters as much as what you type. The tier-by-tier breakdown is in Is It Safe to Put Client Data in ChatGPT?
How do you pull buyer-side comps with AI?
Let a scoring tool build the shortlist, then cut it yourself. RPR launched an AI CMA inside RPR Mobile with a dedicated Buyer CMA mode.
- Scoring: each comp gets a 0–100 score, and the tool flags signals about seller motivation.
- Default filters: same city or ZIP, same property type, sold from the last three months, and square footage within 20% of the subject.
Those defaults are a starting point, not a verdict. Remove the comp that backs to the highway. Remove the flip with finishes your subject doesn’t have. Add the pending sale you know closed over the list. The AI saw more options faster. You still own the comp set.
How do you turn comps into an offer range?
Paste your curated comps into a general assistant with a prompt that forces it to show its work. This is the one I’d save:
You’re helping a licensed real estate agent prepare a buyer-side comparative market analysis. Use only the data I paste below. Don’t invent facts, comps, or numbers.
Subject property (active listing): [beds, baths, sq ft, lot, year built, list price, days on market, price history, listed upgrades]
Comps: [table: comp ID, status, list price, sale price, sold date, days on market, sq ft, beds/baths, condition notes, concessions]
Buyer constraints: [max price, financing type, timeline, required contingencies]
Produce:
1. One paragraph on where the list price sits against the comps, using price per square foot and each sold comp’s list-to-sale ratio.
2. An estimated sale price range (low, likely, high), naming the two or three comps that best support each end. Call it an estimated sale price, never a value.
3. Seller-side signals in the data: days on market versus comps, price reductions, comp concessions.
4. Questions the data can’t answer that the agent should verify before writing the offer.
5. Plain-language explanations a buyer can follow. Describe properties only. Don’t describe the neighborhood’s residents, schools, or safety.
Flag every number you calculated so I can check it.
That last line matters. Language models make arithmetic mistakes. A flagged number takes ten seconds to verify. An unflagged one ends up in your buyer’s offer.
What goes in the “verify before you write” section?
This is the part of the report buyers remember. The AI lists what the data can’t tell you, such as:
- why a comp sold $20,000 under list
- whether the subject’s “updated kitchen” means 2024 or 2014
- whether the seller already bought their next home
Each item becomes a call to the listing agent, a line on the inspection list, or a question for the lender. It turns a report into a plan.
How do you package it for the client?
Keep it to one page, plus a comp table. If you practice in Texas, add the required TREC statement before it goes anywhere (see the compliance section below).
Build the template once inside a Claude Project so the structure, disclaimer, and voice are already loaded every time. The setup is in What Is a Claude Project, and How Do Agents Use One?
What compliance lines apply to an AI buyer report?
Three apply.
The TREC statement. Under 22 TAC §535.17, a Texas license holder can’t give an opinion of value without an appraiser’s license. Any CMA or estimated sale price must include TREC’s required statement:
- verbatim
- in at least 12-point font
- in the sponsoring broker’s name when a sales agent prepares it
The rule covers the person the analysis is prepared for. That includes your buyer.
Fair housing. NAR’s guidance on why every brokerage needs an AI use policy names neighborhood descriptions that drift into steering as a specific AI risk. It also warns that AI can’t determine property value without human oversight. Buyer reports are where this shows up, because AI likes to add “context.” Describe the property, and point buyers to original public sources for anything else. The prompt-level fix is covered in AI Real Estate Ads and Fair Housing.
MLS data terms. Before you paste an MLS export into any AI tool, check your MLS’s data license and your brokerage’s AI policy.
This is general information, not legal advice. Confirm your practices with your broker or an attorney.
Common mistakes
- Treating the on-market Zestimate as a check on list price. It’s partly derived from the list price. Show the comps.
- Letting the AI choose the offer. It can draft scenarios. The offer is your buyer’s decision, made with your advice.
- Skipping the math check. Ask the model to flag every number it calculated, then verify them.
- Adding the neighborhood “color.” Schools, safety, and who a neighborhood is “great for” don’t belong in an AI-drafted report.
- Dropping the TREC statement because it’s “just for my buyer.” The rule applies to anyone the analysis is prepared for.
Frequently Asked Questions
Can AI build a CMA for a buyer?
Yes. AI CMA tools can pull and score buyer-side comps, and a general assistant like Claude or ChatGPT can turn that comp set into a plain-language offer analysis. RPR’s AI CMA inside RPR Mobile includes a dedicated Buyer CMA mode. The agent still selects the final comps, checks the math, and advises on the offer, because the tool can’t walk the property or confirm what the data suggests.
What should be in a buyer CMA report?
A strong buyer report has five parts: the subject property’s facts and price history, a curated comp set with price per square foot, an estimated sale price range, seller-side signals like days on market and price reductions, and a list of open questions to verify before writing the offer. In Texas, it also carries the TREC estimated-sale-price statement.
Can I paste MLS data into ChatGPT or Claude?
Check two things first: your MLS’s data license terms and your brokerage’s AI policy. MLS agreements govern how exported data can be reused, and some consumer AI accounts use your inputs for model training by default. Use a business-tier account, paste only the fields the analysis needs, and never include client names or financial details. This is general information, not legal advice.
Do Texas agents need a disclaimer on a buyer CMA?
Yes. Under 22 TAC §535.17, any comparative market analysis or estimated sale price a license holder provides must include TREC’s required statement that it represents an estimated sale price and isn’t an appraisal. The statement must be reproduced verbatim in at least 12-point font, and a sales agent submits the analysis in the sponsoring broker’s name. This is general information, not legal advice.
Should a buyer report include the Zestimate?
Include it only to explain it. Zillow reports a 1.9% median error for homes on the market, but it also lists the listing price among its inputs, so an on-market Zestimate isn’t fully independent of the number you’re trying to test. Show the buyer the comps instead, and use the estimate as a talking point, not a benchmark.
Can AI tell my buyer what to offer?
No. AI can organize the comps, calculate price per square foot, and draft offer scenarios, but the offer is the buyer’s decision made with the agent’s advice. The model can’t see the property’s condition, hear what the listing agent said, or weigh how badly this buyer wants this house. Use AI to build the analysis and keep the recommendation human.
How long does an AI buyer report take?
Drafting takes minutes once the system is built. Comp scoring in an AI CMA tool and a saved narrative prompt handle the assembly; checking the comps and the math is the part you don’t skip. Build the template once in a Claude Project or Custom GPT so every report starts from the same structure.
Bring this to your team or event
Emily Terrell speaks at brokerage events, real estate conferences, and team trainings on AI, systems, and social media — the exact playbook in this post, delivered live to your audience. As a Top Coach and Speaker at Tom Ferry International and an active agent closing 70+ transactions a year, Emily speaks from the stage about what’s working right now, not theory. Recent stages include NAHREP and eXp Con.
Book Emily to speak at your next event:
Email: eterrell@yourcoach.com
Phone: (210) 400-9191
Web: coachemilyterrell.com
For real estate agents who want to implement this: Get the weekly real estate prompt library at weeklyrealestateprompts.com or follow @coachemilyterrell on Instagram for daily systems and AI breakdowns.