Can ChatGPT Do a CMA? What General AI Gets Wrong About Comparative Market Analysis
Agents ask this every week, usually right after a general AI tool produced something that looked like a CMA and turned out to be unusable in front of a seller. The honest answer separates two very different jobs: formatting a CMA and sourcing one.
The short answer
ChatGPT can format a CMA. It cannot source one. A general-purpose AI tool has no connection to your MLS, so every comparable it produces comes from what you paste in or from public data it has absorbed, which may be incomplete, out of date, or simply wrong. It will still produce a confident, well-organized document. That is the risk.
If you want AI to help with pricing work, the question is not which model writes the best summary. It is which tool is connected to the data the summary depends on.
What a CMA actually requires
A defensible comparative market analysis is a chain of judgments, and only the last link is writing:
- Pull the right universe of properties, including sold, pending, active, and expired.
- Filter to genuinely comparable properties on location, size, condition, age, and style.
- Verify status and dates, because a comp that closed nine months ago prices a different market than one that closed last month.
- Adjust for differences, such as lot size, finished basement, updated kitchen, view, or a busy road.
- Reconcile the adjusted values into a defensible range.
- Present it in a way a seller understands and believes.
General AI can help with step six. It has no reliable access to steps one through three, and it has no basis for step four beyond generic assumptions.
Three ways general AI breaks a CMA
1. Stale or missing data
Public property records and syndicated listing sites lag the MLS. Status changes, price reductions, and pending activity are the fastest moving signals in pricing work, and they are exactly what public sources report last. An AI answer built on that data can be weeks or months behind the market you are pricing into.
2. Comps that do not exist
When a language model is asked for comparable sales and has no data source, it does what language models do: it produces plausible text. Addresses that look real, sale prices that look reasonable, dates that fit. Presenting a fabricated comp to a seller is not a small error. It is a credibility problem you may not get a second chance to fix.
3. No adjustment logic
Adjustments are local. What a finished basement is worth in one submarket is not what it is worth two towns over, and no general model knows your market’s convention for it. Without MLS data and local context, an AI tool either skips adjustments or invents a rule of thumb.
Where general AI genuinely helps
None of this means agents should keep AI out of pricing work. It means the tool has to sit in the right part of the workflow. General AI is useful for:
- Turning a set of numbers you already trust into clear seller-facing language
- Drafting the narrative around a price recommendation
- Preparing for the pricing conversation, including likely objections
- Explaining a range to a seller who is anchored to a number from a portal estimate
That is real time saved. It is just not the part of the CMA that carries the risk.
Connected AI versus general AI for pricing work
| CMA step | General AI chatbot | MLS-connected AI workspace |
|---|---|---|
| Pull candidate properties | No listing access. Relies on pasted data or public sources. | Queries live MLS data directly. |
| Verify status and dates | Often stale or unavailable. | Current status, days on market, and price history. |
| Select comparables | Generic similarity, no local judgment. | Ranked against the subject property on real listing attributes. |
| Adjustments | Rules of thumb at best. | Grounded in what comparable properties in that market actually did. |
| Seller-ready output | Strong writing, unverified numbers. | Strong writing on numbers you can trace back to the source. |
| Agent review required | Yes, on every number. | Yes, on judgment and positioning. |
Note the last row. Connected AI does not remove the agent from pricing. It moves your review from data checking to judgment, which is the part clients are actually paying you for.
How REX approaches it
REX is an AI workspace built for real estate rather than a general assistant pointed at real estate. With MLS Connectivity enabled, REX works from live MLS data, so comparable research, property history, and market context come from the same source your pricing decision does. From there it does the drafting work too: the CMA narrative, the seller-facing explanation, and the objection prep for the listing appointment.
The practical difference is where your time goes. Instead of assembling data and then writing about it, you review a draft and spend your attention on strategy and the conversation.
A checklist before you present any AI-assisted CMA
- Confirm every comparable exists in the MLS and pull it up yourself.
- Check status and closing date on each one.
- Confirm the adjustments reflect your market, not a national average.
- Confirm nothing material is missing, including pending sales and recent price changes.
- Read the narrative for anything you would not say out loud to a seller.
If a tool cannot survive that checklist without rework, it is not saving you time. It is moving the work later in the process, to the point where mistakes are most expensive.
Frequently Asked Questions
Can ChatGPT do a CMA?
ChatGPT can format and write a CMA, but it cannot source one. It has no MLS access, so any comparable sales it produces come from data you paste in or from public sources that lag the MLS. It can also generate comparable properties that do not exist. Use it for drafting language around numbers you have already verified, not for selecting comps.
Can AI connect to the MLS?
Some real estate specific AI tools can, subject to your MLS rules and approvals. General assistants such as ChatGPT, Gemini, and Claude cannot. MLS connectivity is what separates an AI tool that can do comparable research from one that can only write about it. REX offers MLS Connectivity as an add-on to any plan.
Is it safe to use AI for pricing?
It is safe when the data is sourced from the MLS and the agent reviews the result. It is not safe when the tool has no data connection, because the output looks equally confident either way. Always verify comparables against the MLS before presenting a price to a client.
How much time does AI actually save on a CMA?
The savings come from research and drafting rather than from judgment. Pulling and organizing comparable data and writing the seller-facing narrative are the repetitive parts. Selecting comps, setting adjustments, and defending the range still require you. Agents who see the biggest gains use AI for the first group and keep their attention on the second.
What is the difference between an AI CMA and an automated valuation model?
An automated valuation model produces a single estimated value from an algorithm, which is what portal estimates do. An AI-assisted CMA is a working document built from selected comparable properties with adjustments an agent can explain and defend. Sellers can argue with an estimate. A CMA gives you something to walk them through.
The takeaway
The question is not whether AI belongs in pricing work. It is whether your AI tool is connected to the data that pricing work depends on. A tool that writes beautifully from bad comps is a liability in a listing appointment. A tool connected to the MLS gives you back the hours without putting your credibility on the table.
Learn more about REX solutions, start a 30-day free trial or talk with our team about rolling it out across your office.
