Listing Descriptions with ChatGPT: 5 Prompts That Actually Work (With Examples)
Most ChatGPT listing prompts produce generic American-flavoured slop. These five are tuned for AU markets and produce copy your VPA team will actually run. Real examples included for each.
The default ChatGPT listing description is recognisable from across the room. "Nestled in the heart of...", "an opportunity not to be missed", "boasting an array of features." Every agent who's tried using ChatGPT for listings has wasted an afternoon trying to make those clichés go away.
The fix isn't a better AI; it's a better prompt. These five prompts give ChatGPT enough constraint to produce copy that sounds like a competent AU agent wrote it, in 30 seconds.
Why the default ChatGPT output fails
ChatGPT was trained on the internet, which is mostly US English real estate copy. Without explicit instruction, the model defaults to American spelling, US property terminology ("yard" not "backyard"), and US-style superlatives ("absolutely stunning"). It also leans heavily on three cliché patterns:
- The "nestled in" opener
- The "boasts a..." feature list
- The "opportunity not to be missed" close
Stripping these requires telling ChatGPT exactly what to avoid AND what to do instead. That's what these prompts do.
Prompt 1: The voice-matched listing description
This is the workhorse prompt. Provide 2 to 3 of your past listings as voice samples, list the new property's features, and ChatGPT writes in your existing style.
Four bedrooms upstairs include a master with walk-in robe and ensuite. The remaining three share a renovated family bathroom. Downstairs, a separate living area opens to a fully fenced 612sqm block with established gardens, mature trees, and side access to the double garage.
Tight-held streets in this part of suburb rarely come to market; when they do, families upgrading from townhouses in the area tend to move fast. Walking distance to the local primary, two minutes to the main café strip, fifteen minutes to the CBD.
Prompt 2: The buyer-first opener variant
For listings where the buyer story matters more than the architecture (typical of established family homes), open by describing the lifestyle rather than the building.
The bones support the lifestyle. Four bedrooms upstairs, a renovated family bathroom plus master ensuite, and a separate downstairs living area that doubles as the teen retreat or the home office. The kitchen was redone last year with stone and integrated appliances.
Tight-held streets in this pocket rarely surface to market. When they do, the buyers who win are usually upgrading from a townhouse nearby and have been watching this exact block for two years.
Prompt 3: The 90-word social/MMS variant
For database alerts, SMS, or Instagram captions. Tight, hook-forward, no fluff.
Prompt 4: The "tell me what's unsaid" prompt
For properties with quiet but important features (the orientation, the school zone, the quiet street), this prompt makes ChatGPT lead with the thing the buyer is actually buying.
This one is useful when you have 12 to 15 features and you can't tell which to lead with. ChatGPT's first response will name the spine; if you disagree, edit and re-prompt. Saves 15 minutes of "what's the angle on this one?".
Prompt 5: The price-range comparison prompt
For properties where the price-range is the unspoken question (and it usually is), this prompt produces copy that anticipates the comparison.
This prompt produces copy that does the work the buyer does silently. They're not asking "is this house good?", they're asking "is this house better than the three alternatives I'm currently considering?" The description should answer the second question.
Common mistakes
Even with the right prompts, three mistakes I see most often:
- Skipping the voice samples. Without 2 to 3 of your past listings as reference, ChatGPT defaults to generic. Always include samples.
- Accepting the first draft. Spend 2 minutes editing. Strip any remaining clichés. Add one specific local detail (the café you actually go to, the school your buyer's kids will attend). That's what makes it not-look-like-AI.
- Using the same prompt for every listing. Apartments and family homes need different prompts. Use Prompt 1 as the baseline, swap to 2 to 5 based on the property.
What this looks like at scale
For an agent doing 10 listings a month, the math:
- Before: 25 minutes per listing × 10 listings = ~4 hours/month on listing copy
- After: 3 minutes per listing (1 min ChatGPT + 2 min edit) × 10 = 30 minutes/month
- Saved: ~3.5 hours/month per agent, ongoing forever
Multiply by 3 agents in a typical boutique agency and you're at ~10 hours/month back. Not the biggest of the 6 highest-ROI automations, but the easiest to ship.
FAQ
Can ChatGPT write good real estate listing descriptions?
Yes, with the right prompt. The default ChatGPT output produces generic American-flavoured copy with cliches. With a tuned prompt that includes voice samples, market context, and Australian English, the output is publication-ready in 30 seconds with a 2-minute edit.
What's the best ChatGPT prompt for real estate listings?
The 'role + voice + constraints + property data + word limit' framework works best. Tell ChatGPT it's writing for an AU sales agent, provide 2 to 3 voice samples from past listings, list the property's specific features and target buyer, set Australian English as a constraint, and cap the output at 180 to 220 words.
Will buyers know my listing description was written by AI?
Probably not, when you edit the AI draft for 2 minutes per listing. The dead giveaways are American spelling ("color" instead of "colour"), cliches like 'nestled in', and corporate phrases like 'an opportunity not to be missed'. Strip those, add a specific local detail, and the description reads as written by a competent local agent.
Should I use ChatGPT-4 or Claude for listing descriptions?
Both work well. ChatGPT-4 is slightly better at voice matching when you provide samples. Claude is slightly better at flowing narrative paragraphs. For agents doing this manually, either is fine. For agencies integrating into a CRM workflow, ChatGPT's API is more widely supported in tools like Zapier and Make.com.
How long does it take to write a listing description with ChatGPT?
30 seconds for the AI to generate, 2 minutes for you to edit. Total: about 3 minutes per listing, compared to 20 to 30 minutes manually. For an agent writing 10 listings per month, that's roughly 3.5 hours saved per month, ongoing.