Starbase / 01 ListingPrompt

ListingPrompt / Guides

Real estate field guide

AI listing description prompt

Most AI listing descriptions read the same: 'nestled', 'boasts', 'stunning'. The problem is rarely the tool. It's the prompt. If you paste three bullet points and ask for 'a listing description', the model fills the gaps with clichés and, worse, with features the home doesn't have.

The prompt below is the one from our 'Listing description in 5 minutes' playbook. It works in ChatGPT or Claude. It forces the model to lead with the strongest feature, use concrete details, stay clear of Fair Housing problems, and hand you headline options you can test.

The prompt

You are a real estate copywriter. Write a 150-word MLS listing description for: [beds/baths/sqft/address/features]. Lead with the strongest feature, use concrete sensory details, avoid Fair Housing violations (no mention of who the home is 'ideal for' by family status, religion, etc.), and end with a call to schedule a showing. Then give 3 headline options under 60 characters.

Step by step

  1. Collect the facts first: beds, baths, square footage, lot, year built, updates with years, and the two or three things buyers comment on at showings.
  2. Paste the facts into the [brackets]. Use plain notes; the model does not need full sentences.
  3. Read the draft against your facts line by line. Delete anything you can't verify.
  4. Check your MLS character limit and ask for a shorter version if needed ('rewrite in under 900 characters').
  5. Pick one of the three headlines for the portal title and save the others for social posts.
SPONSOR SLOT · AVAILABLEFeature your product here

What makes this prompt work

It sets a role (copywriter), a length (150 words), a structure (strongest feature first, call to action last) and a constraint (Fair Housing). Those four things are what separate usable copy from filler. The headline request gives you options for portals and social in the same pass.

'Concrete sensory details' pushes the model away from adjectives and toward specifics, such as morning light in the kitchen or the sound of a creek out back, but only if you supplied those details. If you didn't, it will invent them, which is why the fact check matters.

Make the output sound like you

Paste two of your past descriptions you liked and add: 'Match the tone and sentence length of these examples.' Ask for a second version that is plainer and more factual, then merge the best lines. If the draft still feels generic, tell the model which word to ban: 'Do not use nestled, boasts, stunning or oasis.'

Mistakes to avoid

Describing people instead of the property ('perfect for a young family'). Describe the home and its features instead.

Letting the model add schools, distances or commute times you didn't give it. Those are claims you are responsible for.

Publishing without checking the MLS character limit, which can cut your call to action off mid-sentence.

Save it as a reusable assistant

Retyping prompts is where most agents give up. In ChatGPT, create a project or custom GPT for this task, paste the prompt into its instructions, and add one or two examples of output you were happy with. In Claude, a project with the same instructions does the job. From then on you only paste the facts. Keep a short note of what you changed each time you edit the instructions, so you can go back if a change makes the drafts worse.

A two-minute review checklist

Before anything AI-drafted leaves your hands, run through the same five checks. Is every fact, number and feature in the draft something you supplied or can verify? Does it describe the property and the service rather than the kind of person who should buy? Is it the right length for where it's going: MLS field, text message, email or caption? Does it sound like you, or like an ad? And does it include anything your brokerage or state requires, such as brokerage name, license number or a staging disclosure? If any answer is no, fix it or ask the model for a revision with that specific instruction.

Keep a running list of the fixes you make most often. After a week or two, add them to the prompt itself. That's how a generic prompt becomes your prompt, and it's the single biggest improvement you can make to AI output.

Before you use it

Every prompt on this site is a starting point, not a finished product. AI drafts fast, but you are the licensed professional: check every fact against the MLS and your own notes, keep Fair Housing rules in mind, and follow your brokerage's and your state's advertising rules. Treat the output as a first draft from a quick junior assistant who has never seen the property.

Frequently asked questions

Can ChatGPT write MLS listing descriptions?
Yes, as a first draft. Give it only verified facts, set the length, and read every line before it goes on the MLS. You remain responsible for accuracy and compliance.
How do I keep AI listing copy Fair Housing compliant?
Tell the model to describe the property, not who should live there, and review the result yourself. Avoid references to family status, religion, race, disability or other protected classes. When unsure, ask your broker.
ChatGPT or Claude for listing descriptions?
Both work with this prompt. Try the same facts in each and keep whichever voice is closer to yours.

More guides

Back to all of ListingPrompt

Newsletter

Get one new realtor playbook every week

Free. No spam. Unsubscribe anytime.