AI Is Writing Most Listing Descriptions. Who Checks Them?
AI·November 17, 2026
Three out of four real estate agents who use artificial intelligence say they rely on it to write listing descriptions, according to 2026 data from the National Association of Realtors. That is a striking share for a tool that was a novelty in brokerage offices only a few years ago. It also raises a practical question: when a machine drafts the words that sell a home, who makes sure they are accurate?
The line between marketing and misrepresentation has never been especially sharp. Listing copy has always leaned on words like "charming," "move-in ready," or "boutique," and buyers have learned to read past them. AI tools make that kind of promotional language faster and cheaper to produce, but they also make it easier to overstate. A model asked to make a property sound appealing may describe a "renovated kitchen" without knowing what was actually updated, or call a basement "bonus living space" even when it doesn't meet permit or egress standards.
The real risk sits less in the software than in the review step. An agent who approves generated text is still the one answerable for it. Obligations under state licensing rules, MLS policies, and fair housing law do not change because a draft came from a chatbot. Brokers who want to limit exposure tend to set simple rules: verify every factual claim against the property record, keep descriptions consistent with photos, and avoid language that could steer buyers by neighborhood, family status, or other protected characteristics. That last point is one automated writing can get wrong without anyone intending it.
For buyers and sellers, the takeaway is that the listing they read may have started as a model's draft and received only light editing. Checking disclosed square footage, features, and permit history against public records, and asking the agent direct questions about anything unusual, are sensible habits in that setting.
The headline figure also has limits. It covers agents who already use AI, so it says nothing about how many agents use it overall or how much generated text goes out unedited. Still, it shows how quickly the technology has moved from experiment to routine in a part of the business where a single inaccurate sentence can carry legal and financial consequences.
Reporting based on an external source.