Using AI for Real Estate Listing Descriptions Without Fair Housing Problems

Yes, you can use AI for real estate listing descriptions, and it can save real time on first drafts. The safe way to do it is simple: build a facts sheet first, let the AI write only from those facts, check every line against the file, then run a fair housing and California disclosure pass before anything goes to the MLS. The California Department of Real Estate put it plainly in a March 2026 advisory on AI: if an AI tool produces misleading advertising, responsibility rests with the licensee and the responsible broker, "not the technology provider."

Why the facts sheet comes before the prompt

Plenty of agents already use these tools. In the National Association of Realtors 2025 Technology Survey (an association survey of 1,241 members), 20% said they use AI daily, while 32% had not actively tried it for their business. On impact, 46% reported neutral or no noticeable impact. So the tool is common, but it is not magic.

The biggest problem with AI listing copy is not style. It is that a language model fills gaps. Ask for "a warm description of a 3 bedroom home in San Rafael" and it may add hardwood floors, an updated kitchen or a quiet street that nobody confirmed. The fix is to give it nothing to guess about.

A good facts sheet has:

• Beds, baths, and square footage with the source (assessor record, appraisal or measurement)
• Lot size and any known easements or shared driveways
• Year built and what has actually been updated, with dates the seller confirmed
• Features you saw yourself on the walkthrough
• Anything the seller asked you not to mention
• Your MLS character limit for public remarks

A step by step workflow for AI listing descriptions

1. Fill the facts sheet from the disclosures, the public record and the seller's confirmed answers. Nothing goes in from memory.
2. Use a business account, not a personal one, and leave out seller names, phone numbers, alarm codes and anything private. The DRE advisory tells licensees to avoid entering confidential client information into public or unsecured AI platforms.
3. Paste the prompt below with the facts sheet underneath it, in ChatGPT or Claude.
4. Ask for two versions, one for the MLS limit and one longer for your website.
5. Check every sentence against the facts sheet. Delete anything you cannot point to.
6. Run the fair housing pass described below.
7. Add the California disclosures that apply, then have your broker or transaction coordinator give it a second read. The DRE recommends keeping records of how AI output was reviewed.

A prompt you can copy

You are helping a California real estate agent write MLS public remarks. Use only the facts listed below. Do not add features, condition claims, school claims, or neighborhood claims that are not in the list. Describe the property, not the kind of person who should buy it. Keep it under [your limit] characters, plain and specific, without phrases like "nestled" or "boasts." After the draft, list every factual claim you used so I can check it.

That last line matters. It turns your fact check into a quick checklist instead of a reread.

Fair housing rules that apply to AI written listings

The federal rule is short. The Fair Housing Act, 42 USC 3604(c), makes it unlawful to publish any notice, statement or advertisement for the sale or rental of a dwelling "that indicates any preference, limitation, or discrimination" based on race, color, religion, sex, handicap, familial status or national origin. HUD's regulation, 24 CFR 100.75, says this covers "words, phrases, photographs, illustrations, symbols or forms," and also covers selecting media or locations for ads that deny particular groups information about housing. That second part is worth remembering if you use automated ad targeting. HUD's own Fair Housing Act overview lists the protected groups.

California goes further. Government Code section 12955 applies the same advertising rule to a longer list that includes gender identity, sexual orientation, marital status, ancestry, source of income, veteran or military status and genetic information. The DRE advisory also points to its own regulation, section 2780, on discriminatory advertisements.

In practice, AI drafts drift into describing buyers. Lines like "perfect for a young family," "ideal for empty nesters" or "a great fit for a couple" describe people, not the property. Our rule for every draft, even when a line is not clearly unlawful: strike any sentence about who should live there and replace it with a fact about the home, such as the number of bedrooms, the yard or the distance to transit.

California rules: license numbers and altered photos

Two California statutes catch AI marketing.

License disclosure. Business and Professions Code section 10140.6 requires a licensee to disclose their name, license identification number and responsible broker's identity on all solicitation materials intended to be the first point of contact with consumers. The DRE advisory says this still applies when the content is AI generated. If you use AI to produce postcards, social posts or landing pages, build those fields into the template so they are never left off.

Digitally altered images. AB 723 (Chapter 497, Statutes of 2025) added Business and Professions Code section 10140.8. The DRE says licensees must comply as of January 1, 2026. If an ad for the sale of real property includes a digitally altered image, it must carry a reasonably conspicuous statement, on or next to the image, that the image was altered, plus a link, URL or QR code to the original, unaltered image. On a website you control, you must include the unaltered version (a link to it also works).

The law defines an altered image as one changed with photo editing software or artificial intelligence to add, remove or change elements, including furniture, fixtures, flooring, paint color, landscape, views through windows and neighboring properties. Virtual staging and AI decluttering fall in that list. Ordinary adjustments to lighting, white balance, color correction, cropping, straightening or exposure that do not change how the property is represented are excluded.

Rules and MLS policies change, so confirm your listing workflow with your broker or a real estate attorney before you rely on it.

Are AI listing descriptions obvious?

Often, yes. Unedited drafts lean on the same words ("stunning," "boasts," "nestled," "oasis") and the same three sentence rhythm. Buyers and other agents notice. The cure is the facts sheet plus one or two details only you know from the walkthrough, such as where the afternoon light lands or how the garage is set up for storage. AI is good at structure and trimming to a character limit. Your observations are what make it read like a listing instead of a template.

What to watch out for

• Invented features. The DRE advisory cites section 10176 on substantial misrepresentation and says use of an AI tool "will not excuse the violation." Anything not on the facts sheet comes out.
• School and neighborhood claims. Do not let AI name school assignments or describe a neighborhood's character. Point buyers to the source instead.
• Photo edits without the disclosure. Ask your photographer and staging vendor which images are altered, and keep the originals in the listing folder.
• Ad targeting. Automated audience choices can exclude groups even if your copy is clean.
• Private data. Seller details, lockbox codes and offers stay out of AI tools.

The same discipline carries over to rentals; our guide to AI for property managers covers tenant ads and screening, and responding to Google reviews with AI covers the other place agents write in public every week.

Where to start

Build a facts sheet template this week and run your next listing through the seven steps above, timing how long it takes. If you want help setting up templates and a review checklist for your whole office, see our AI consulting services or our AI assessment, or book a free 15 minute call.