Six months ago, half the agents I know had never heard of ChatGPT. Now I can’t have a coffee meeting without someone showing me their latest prompt engineering technique.

I spent the past month interviewing 15 high-performing agents across Sydney about how they’re actually using AI writing tools. The range of approaches—from sophisticated workflows to cringe-worthy shortcuts—tells us a lot about where this industry is heading.

The Good: Streamlined First Drafts

The most effective agents are using ChatGPT to generate first drafts that they then heavily edit. It’s not replacing their voice—it’s handling the blank-page problem.

Rachel, an agent in Mosman, described her workflow: “I walk through the property with voice notes, recording every feature and impression. Then I feed those notes to ChatGPT and ask for a draft description. It gives me something to react against, which is faster than starting from scratch.”

Her key insight: the AI output is never the final product. It’s a starting point that captures the basic information while she adds the persuasive flourishes that actually sell properties.

Another agent in the Inner West uses ChatGPT specifically for the mundane parts of listings—the standard features, the logical floor plan descriptions, the suburb boilerplate. He reserves his creative energy for the opening hook and the unique selling points.

This hybrid approach respects what AI does well (structure, comprehensiveness) while preserving what humans do better (persuasion, emotional resonance).

The Bad: Copy-Paste Without Review

Then there’s the other end of the spectrum.

I’ve seen listings appear on the market with obvious AI artifacts. Phrases like “this stunning property offers” (ChatGPT’s favourite opener) repeated across multiple listings from the same agency. American spellings that slipped through. References to features the property doesn’t have because the AI hallucinated them.

One particularly bad example in Parramatta described a property as having “breathtaking views of the harbour”—despite being 25 kilometres inland with no water views whatsoever.

This isn’t just embarrassing. It’s potentially misleading under NSW fair trading laws. If your AI-generated description misrepresents a property, you’re liable, not the AI.

The agents doing this are optimising for speed over quality. They’ll learn the hard way when a vendor complains about an inaccurate listing or when buyers waste everyone’s time inspecting properties that don’t match the description.

The Ugly: When AI Meets Vendor Expectations

Here’s a scenario I’ve encountered multiple times now.

Agent uses ChatGPT to write glowing property description. Vendor reads description and gets unrealistic expectations about interest and price. Property goes to market and underperforms relative to the inflated copy. Vendor blames agent for the gap between description and reality.

The AI is very good at writing enthusiastically. It’s trained to be helpful and positive. This can backfire spectacularly when vendors expect the description to become reality.

One agent told me about a difficult vendor conversation after an AI-written description called their property “the finest example of Art Deco in Randwick.” The vendor then rejected offers below what other “finest examples” had achieved—even though the property was actually fairly standard for the area.

Smart Prompts I’ve Collected

The agents getting real value from AI have developed sophisticated prompts. Here are some approaches worth borrowing:

The Constraint Prompt “Write a property description for [address]. Maximum 150 words. Do not use the words ‘stunning’, ‘immaculate’, or ‘must-see’. Include these specific features: [list]. Target buyer: downsizing couple from [nearby suburb].”

Constraints force the AI away from generic output.

The Comparison Prompt “Here is a listing I wrote that sold well: [example]. Write a new listing for [different property] in a similar style and tone.”

Teaching the AI your voice through examples works better than trying to describe your style.

The Objection-Handler Prompt “This property has been on market for 45 days with limited interest. Current description: [paste]. Identify what’s weak about this description and rewrite to address likely buyer objections.”

Using AI to diagnose problems, not just generate content.

The Workflow That Actually Works

Based on my interviews, here’s the approach I’m recommending to agents I consult with:

1. Property inspection notes: Voice record your genuine impressions during the walkthrough. Include negatives and challenges, not just positives.

2. AI first draft: Feed your notes to ChatGPT with specific constraints (word count, banned phrases, target buyer).

3. Manual edit pass: Cut the AI’s generic flourishes. Add specific details only you know. Ensure accuracy.

4. Vendor review: Have the vendor read it for factual accuracy before publication. This catches AI hallucinations and sets realistic expectations.

5. Performance tracking: Note which listings generate strong enquiry. Feed successful examples back to the AI.

This workflow takes about 20 minutes per listing—roughly half the time of writing from scratch, but with quality control built in.

What This Means for Days on Market

Here’s the business case for getting this right. According to REA Group data, listings with comprehensive descriptions receive significantly more engagement. Agents who write compelling, accurate descriptions see measurably better outcomes. Buyers who arrive at inspections with accurate expectations convert at higher rates. Properties with engaging descriptions spend fewer days on market.

The AI can help you get to good copy faster. But only if you maintain quality control. Slapping ChatGPT output directly onto the portals is a false economy that will hurt your clearance rates and vendor relationships.

The Future of AI in Real Estate Marketing

I expect AI writing tools to get significantly better over the next year. We’ll see integration directly into CRM systems and listing platforms. The technology will learn Australian English and real estate conventions.

But the fundamental principle won’t change: AI is a tool, not a replacement for professional judgment. The agents who thrive will be those who use it to handle routine work while focusing their human creativity on what actually wins listings and closes sales.

The worst thing you can do is nothing—either ignoring AI entirely or using it thoughtlessly. Find the middle path, and you’ll have an advantage over both camps.

---

Linda Powers advises real estate agencies on technology adoption and marketing strategy. Her 25 years in Sydney real estate included pioneering early internet marketing techniques in the 2000s.