How to Use ChatGPT for Real Estate Listings: Real Cases

How agents use ChatGPT and other AI for listing descriptions

The common workflow has three steps:

1. Collect the facts. The agent gathers notes, the MLS sheet or a showing recording.
2. Get a draft. The AI writes a first version.
3. Edit it. The agent rewrites it in their own voice.

What changes from case to case is the input and how much the agent rewrites.

6 records cover this task. Not counting the 1 photo caution, they come from 5 people. 5 name ChatGPT and 1 name Claude.

Tools named, per record (listing descriptions)

1. ChatGPT5 of 6
2. Minutes AI1 of 6
3. Claude1 of 6

The current cases show three variants:

• Notes and the MLS sheet into a saved setup. Brad Cohen started by pasting his listing-appointment notes and the MLS sheet into ChatGPT. Later he and his operations manager built a custom GPT. It takes one listing and returns the description plus social posts, an email blast and a video script. He calls it seven deliverables from one input (Brad Cohen).
• Recorded showing notes into a “story”. The List Realty’s broker-owner walks the home with the seller and records everything with Minutes AI. Then ChatGPT looks for the thread that ties the home together. She says the tool does not write the listing; it only removes the blank page (Broker-owner, The List Realty).
• A role prompt aimed at one buyer. On her first day with Claude, @staceywalterss told it to act as the listing agent and write for a young family, “warm and specific”. She saved the prompt to reuse. This was one hour of use, not a tested workflow (@staceywalterss).

Two of the current workflows start from the agent’s own notes: what they saw at the appointment or the showing (Brad Cohen; Broker-owner, The List Realty). The 2023 agent pasted a Zillow page and says the more detail you give, the better the output (@kalachejackson_realtor).

What they report

Few records give a number, and every number is self-reported. 2 of 6 records state one.

Reported results (listing descriptions, self-reported)

1. Brad Cohen before: by hand~30 min
2. Brad Cohen after: ChatGPT → custom GPT~2 min
3. @kalachejackson_realtor after: ChatGPT (2023) · before: not stated<30 s

• Brad Cohen: ~30 min → ~2 min per description; 7 listing deliverables in minutes
• @kalachejackson_realtor: Draft in under 30 seconds

The strongest number comes from Brad Cohen: about 30 minutes per description before, about 2 minutes after his edits. He also says the full set of listing material now takes minutes, not hours (Brad Cohen).

He speaks as a local business owner and does not sell a course or tool. That makes his the cleanest record on this page, but it is still one person’s estimate.

The other number is from 2023: a draft in under 30 seconds (@kalachejackson_realtor). That measures how fast the tool writes, not how long a usable, checked description takes.

The rest give no number:

• @staceywalterss says her first Claude draft beat half of her own listings already on the MLS (@staceywalterss).
• The List Realty gives a time saving only for meeting notes, which is a different task (Broker-owner, The List Realty).

Why AI listing descriptions go wrong

The cases name four problems. Two are about the words, one is about the facts, and one is about photos.

It does not sound like you

Brad Cohen says the first draft “wasn’t exactly my voice”, and he edits every output (Brad Cohen). The List Realty rewrites every draft in her own voice: “I am not outsourcing creativity” (Broker-owner, The List Realty).

It can make things up

The List Realty says the tool “does make things up sometimes” (Broker-owner, The List Realty). Check every fact in the draft against your notes and the MLS sheet.

Pretty words can hide the facts

@sandyisyouragent asks agents to stop using ChatGPT for “flowery, adjective-written” descriptions. As a buyer, she has to read to the bottom to learn that a home has a new HVAC or new windows, and often she cannot find the lot size.

On a phone, Zillow shows only the first few lines. So those lines must carry the hard facts and say how the home compares to the neighbourhood (@sandyisyouragent).

She does not sell anything here, and she uses ChatGPT herself for other work (@sandyisyouragent). So her point is about how the tool is used, not a rejection of AI.

• 

  Argues againstTikTok2025-04-04

  @sandyisyouragent — listing descriptions

  The agent asks others to stop using ChatGPT for "flowery, adjective-written" listing descriptions.

  ChatGPT (argues against) · Negative view, no number

AI listing photos: where the line is

This case is about images, not text, but it belongs with listings. @chriswhartonrealtor asked ChatGPT to “improve” real photos from his own listing. The result had new landscaping, lighting and a paved path.

He says he has seen similar images on published listings, and he asks if disclosure is enough. He did not use the altered photos (@chriswhartonrealtor).

How strong is this evidence?

The evidence for these problems is thin: 1 record argues against and 1 is a caution. But they agree with the agents who use AI:

• Current agents: every one who describes a review step edits the draft before it goes out (Brad Cohen; Broker-owner, The List Realty).
• The 2023 record shows no edit (@kalachejackson_realtor).
• The first-day Claude video does not mention one (@staceywalterss).

Then and now

The 2023 record shows the simple pattern:

1. Copy the facts. Take the property facts from Zillow.
2. Paste and ask. Paste them into ChatGPT and type “Write an MLS description for this house.”
3. Regenerate if needed. If the draft is “kind of basic”, click regenerate and take the second one.

The video shows no editing step. The text goes to the MLS or into a social caption (@kalachejackson_realtor).

The current records change three things:

• The input is the agent’s own. Appointment notes and recorded showing notes replace a copied Zillow page (Brad Cohen; Broker-owner, The List Realty). One agent adds a target buyer to a pasted listing (@staceywalterss).
• The setup is saved. A custom GPT and a saved prompt replace a fresh chat for each listing (Brad Cohen; @staceywalterss).
• The edit is a fixed step. Brad Cohen and The List Realty both edit or rewrite before anything goes out (Brad Cohen; Broker-owner, The List Realty).

1 record is from 2023–24 and 5 are from 2025–26, so this is a small sample. Treat it as a direction, not a trend line.

What to take from these cases

These points come only from the records above:

1. Feed it your own notes. The agents with the clearest workflows start from what they saw and heard at the home, not a copied listing page (Brad Cohen; Broker-owner, The List Realty).
2. Put the facts first. Ask for the first two sentences to say what the home is, its key systems and the lot size, before any mood words (@sandyisyouragent).
3. Rewrite, then check. Edit the draft into your voice. Then check each fact against the MLS sheet, because the tool can invent details (Brad Cohen; Broker-owner, The List Realty).
4. Save what works. A saved prompt or custom GPT lets you reuse one input for the description and the rest of the listing marketing (Brad Cohen; @staceywalterss).
5. Keep photos real. Do not let AI “improvements” change what the home looks like (@chriswhartonrealtor).

Know the limits:

• One time saving comes from a single agent, and it is his own estimate.
• Two of the speakers have a business tie: a free playbook download (Broker-owner, The List Realty) and a promoted service (@staceywalterss). 2 records on this page have a business tie.

See every record in the case register and the method in How we collect cases.

Questions agents ask

How do I use ChatGPT to write a listing description?

The simplest record pastes the property facts from Zillow into ChatGPT and types “Write an MLS description for this house” (@kalachejackson_realtor). Two current agents start from their own appointment or showing notes, then edit the draft before it goes out (Brad Cohen; Broker-owner, The List Realty).

Can ChatGPT write listing descriptions in my voice?

Not without help in these cases. Brad Cohen’s first ChatGPT draft “wasn’t exactly my voice”. He later built a custom GPT, but he still edits every output (Brad Cohen). The List Realty rewrites every draft in her own voice (Broker-owner, The List Realty).

Are AI listing descriptions bad?

One agent says the flowery ones are. They push facts like a new HVAC or the lot size below the few lines Zillow shows on a phone (@sandyisyouragent). She uses ChatGPT for other work, so her complaint is about how the tool is used (@sandyisyouragent).

Can you use AI photos in a real estate listing?

One agent says AI-generated photos are allowed if they are disclosed, and he asks if disclosure is enough. We have not checked this against MLS rules. He did not use the altered photos of his own listing (@chriswhartonrealtor).