Three in four agents who use AI write listing descriptions with it. Put a read-through before the MLS

The first draft of a listing now often comes from a machine. This week the National Association of REALTORS® put a number on it.

What the report says

NAR released its 2026 REALTORS® Technology Report on September 22. Among agents who use AI, the top use is writing listing descriptions, at 75%. Social media posts come next at 56%, then emails and follow-up at 52%. Nearly half of agents now use AI daily (23%) or weekly (25%). Only 12% say they do not use it and do not plan to, down from the 32% who had not tried it in 2025.

The reasons are practical. Eighty-one percent said saving time is their main goal with new technology, up from 66% a year earlier. The biggest obstacle was the learning curve, named by 63%, followed by cost at 59%. HousingWire’s write-up of the report adds that this shift raises training and compliance needs for brokers.

A day later, Fortune described how Compass rolled out its AI Assistant to its agents. Compass reports that within weeks of the July launch, about 15,000 agents had generated more than 97,000 conversations with it. That is a company-reported figure. The tool goes beyond drafting: according to Fortune, it can produce a list of prospective clients who may be ready to sell and draft a note to reconnect. Fortune also notes a New York City proposal that would require landlords, brokers and listing platforms to disclose AI-generated or digitally altered rental listing photos and videos. It is a proposal, not a rule, but it shows where attention is going.

Why this matters for a solo agent

A big brokerage can train staff, write a policy and add a review step. A solo agent is the whole review step. When a draft takes thirty seconds, the risk moves from writing to reading. A model will write “newly renovated kitchen” because that is what listing copy usually says, whether or not the seller renovated anything.

Your name and your license are on the listing, not the tool’s. So the draft is not the work. The read-through is.

The seven-point read-through

Print this, or paste it into your listing checklist. Run it on every AI draft before it reaches the MLS, a portal or social media.

1. Match every fact to the input sheet. Beds, baths, square footage, lot size, year built, HOA, parking, schools named. If a number is not on your input sheet or in the public record you pulled, delete it.
2. Strike anything you did not see. Models fill gaps with typical features: “hardwood throughout,” “updated systems,” “quiet street.” Keep only what you walked through or the seller documented.
3. Describe the home, never the buyer. Cut phrases about who the home is “perfect for” and keep to the property, its features and its location facts. Check the draft against your brokerage’s fair-housing guidance. This is education, not legal advice; your broker has the final word.
4. Check the photos against the words. If a photo was virtually staged or edited, follow your MLS and state rules on labeling it, and make sure the copy does not describe furniture or finishes that are not in the house.
5. Confirm the remarks fit the field. MLS rules differ on what may go in public remarks, such as contact details, showing instructions or compensation. Read your MLS’s current rules once and turn them into a short list the draft must pass.
6. Read it in your own voice. Say it out loud. If you would not say it at the listing appointment, rewrite it. Sellers notice when their home sounds like every other home.
7. Get the seller’s sign-off and keep the record. Send the final copy to the seller for approval, and save the prompt, the raw draft and the approved version in the deal file.

A draft that passes all seven is ready. A draft that fails the first two twice in a row means your prompt is short on facts. Give the model the input sheet, not a summary from memory.

A decision rule for what to automate next

The report’s second and third uses, social posts and follow-up emails, carry the same risk in a different shape. Before you hand another task to AI, ask one question: can I check the output in under a minute against something I already have? If the answer is yes, as with a listing draft and an input sheet, automate it and keep the read-through. If the answer is no, as with a pricing opinion or a reply to a negotiation, keep drafting it yourself and use AI only to organize your notes.

What to do this week

• Pick your next listing and time yourself: how long the AI draft took, and how long the seven-point read-through took. That is your real cost per listing.
• Turn your MLS’s public-remarks rules into a five-line checklist and add it to your prompt.
• Ask your broker whether the office has written fair-housing guidance for AI-drafted ads. If not, ask what they want you to follow in the meantime.
• Save your best prompt, the one that uses your input sheet, where you can find it next time.

The tools will keep getting faster. The read-through is what keeps your name worth putting on the sign.