Listing Descriptions & Photo Captions

Welcome to Day 1. I'm Bo Peng, and I built this course because real estate is one of the highest-leverage fields for AI right now — but also one of the most legally sensitive. We're going to start where the time savings are biggest and the risk is most manageable: listing copy and photo captions.

Most agents spend forty-five to ninety minutes per listing on description copy. The first listing you finish today should drop that to under fifteen — without sounding like a robot wrote it.

The Listing Description Prompt That Actually Works

Generic AI prompts produce generic copy. The trick is to give the model your actual data and your actual voice constraints. Here is the structure I want you to memorize: Role + Property facts + Audience + Voice + Constraints + Output format.

Open ChatGPT, Claude, or your tool of choice. Paste this template, fill the bracketed sections, and run it.

You are a senior listing agent writing MLS-ready
description copy.

PROPERTY FACTS
- Address: [123 Maple Ave, Ames IA 50010]
- Beds / Baths: [4 / 2.5]
- Square feet: [2,140]
- Year built: [1998]
- Lot: [0.32 acres, fenced backyard]
- Standout features: [updated kitchen with quartz,
  primary suite on main, finished basement,
  three-car garage, new roof 2024]
- School district: [Ames CSD]

AUDIENCE
Move-up buyers, $450-525k budget, relocating
from Des Moines metro.

VOICE
Warm, specific, no real-estate cliches
("cozy," "must see," "won't last").

CONSTRAINTS — FAIR HOUSING
Do NOT reference race, color, religion, sex,
familial status, national origin, disability,
or any proxy phrase ("safe neighborhood,"
"family-friendly," "walking distance to church,"
"empty nesters," "exclusive," "private community").
Describe the property and its features only.

OUTPUT
1) MLS description, 180-220 words.
2) Three alternate hook sentences.
3) Five short bullet highlights for Zillow.

That single prompt does ninety percent of the work. The reason it produces good copy is that it forces the model to commit to specifics — square footage, the actual upgrade list, the real audience — instead of falling back on stock phrases.

Photo Captions in Bulk

For photo captions, batch them. Don't prompt the model image by image — that wastes time and produces inconsistent voice. Instead, give it the photo manifest in one shot.

Write MLS photo captions for this property.
8-14 words each, descriptive, no superlatives,
no Fair Housing risk language.

Photos in order:
1. Front exterior, dusk, porch light on
2. Foyer with hardwood, wainscoting
3. Living room, fireplace, bay window
4. Kitchen, quartz island, gas range
5. Dining room with built-ins
6. Primary bedroom, tray ceiling
7. Primary bath, double vanity, walk-in shower
8. Finished basement, wet bar
9. Backyard, deck, fenced
10. Three-car garage interior

Return as a numbered list.

The model will produce captions like "Quartz waterfall island anchors the renovated kitchen" instead of "You'll love this beautiful kitchen!" — which is the difference between copy that sells and copy that gets ignored.

Fair Housing Act & NAR Article 10 — Pitfalls You Cannot Skip

AI models will happily generate phrases that violate all of the above unless you tell them not to. Here are the categories that get agents in trouble:

• Familial status proxies: "perfect for families," "great for kids," "not suitable for children," "ideal for empty nesters," "adult community" (unless the property qualifies under HOPA's 55+ exemption and you can document it).
• Religion proxies: "walking distance to St. Mary's," "near the mosque," "Christian neighborhood." You can name a school or a park; you should not name a place of worship as a selling point.
• National-origin / race proxies: "exclusive neighborhood," "private community," "traditional neighborhood," references to ethnic enclaves.
• Disability proxies: "able-bodied," "must be able to climb stairs." Describe the stairs; do not describe the buyer.
• Sex proxies: "bachelor pad," "mother-in-law suite" (use "in-law suite" or "secondary suite" instead).

The two-pass workflow: first pass is the prompt above. Second pass is a compliance review — paste the draft back into the model and ask:

Review this listing description for Fair Housing
Act compliance under 42 U.S.C. §3604(c) and NAR
Code of Ethics Article 10. Flag any phrase that
could be read as a preference, limitation, or
discrimination based on a protected class.
Suggest a compliant rewrite for each flag.

That second pass takes ninety seconds and catches things you and the model both missed on the first try.