A buyer used to open a portal, tick a few boxes, and scroll. Now a growing number open ChatGPT instead and simply describe what they want.

Something like a three-bedroom house near Newtown under $1.4 million, with a study and off-street parking. The listings come back in a map view, and the buyer refines by talking, not filtering.

This is not a preview of something years away. It's live now, and it's worth understanding what it changes about how your listings get found.

We've written before about how AI decides which agents get recommended to sellers. This is the flip side: how it decides which homes a buyer actually sees.

In February, realestate.com.au launched an app inside ChatGPT that lets buyers search its listings in plain language. It followed the portal opening up natural language search to all its users, after a year-long pilot.

REA's chief product officer, Jonathan Swift, put the shift down to changing behaviour, with the strongest uptake among younger buyers. That matters at scale, because more than 12 million Australians visit realestate.com.au each month, so even an early feature reaches a large slice of your buyer pool.

Why property suits a conversation

Property search lends itself to this better than almost anything. Buyers rarely know exactly what they want at the start, so they begin broad and narrow as they go.

REA's own research into AI search bears that out. It found that 59 per cent of AI search sessions start wide, with buyers leading on a lifestyle goal or a commute before they ever name a suburb.

They then refine as they go, with 72 per cent adjusting their search at least once while talking to the AI.

The same research found something more telling still. Around 40 per cent of AI searches asked about architectural style, a property's condition, or its renovation potential, the kind of qualitative detail a dropdown filter has never captured.

Buyers are describing their ideal home in their own words, and expecting the listing to answer.

What this changes for your listings

Here's the practical shift. Your listing is increasingly being read and matched by a machine interpreting a buyer's spoken description, not just filtered by a human clicking checkboxes.

If a buyer asks for a single-level home with a north-facing yard and a study, the AI goes looking for listings that clearly say so. The listing that spells those things out surfaces, while the one that leaves them vague can be passed over, even when the home is a perfect fit.

None of this replaces the portal or the filter, and REA is clear on that point. Mr Swift describes the feature as "a partner, not a replacement, for search filters."

It's another front door, and an increasingly popular one.

Where the portals are up to

REA is driving the pace, and buyers now meet its conversational search in two places rather than one. Alongside the ChatGPT app, the portal has rolled its own AI assistant out to all users across web and mobile, so natural language search is becoming native to realestate.com.au itself.

Voice-based AI has also been answering buyer questions on individual listings through providers like Propic since last year, used by agencies including McGrath.

Domain has been slower to launch conversational discovery of this kind, so this is still early. Early is exactly when it pays to get your habits right, before the behaviour becomes the norm.

What stays valuable

It's worth being clear-eyed about what AI does and does not do here. As Elite Agent's Samantha McLean has argued, when the machine handles the search, the suburb comparison and even the first look, the bar for that first meeting with an agent only rises.

What it cannot do is read a street, judge how a home will show on the day, or guide a nervous vendor through the biggest financial decision of their life. Curation, local knowledge and relationships are still yours.

The machine can surface the listing. It still takes an agent to win it and sell it.

Four ways to make your listings work in AI search

The good news is that this is mostly about discipline, not new skills.

Fill every field

Conversational search matches on structured data, so a blank price guide or an empty parking field means the listing will not appear when a buyer asks for those things.

Write descriptions that answer real questions

Buyers are leading with lifestyle and commute before asking about the study, the single-level layout or the north-facing living, so name the station, the school catchment and those features plainly rather than leaning on vague phrases like nestled in a leafy enclave.

Be specific rather than embellished

The AI takes your words literally and matches them to what the buyer actually asked for, so precision brings the right people through the door and overselling brings the wrong ones.

Think in sentences, not keywords

The query is now a full spoken request, so the listing that reads like a clear, factual answer to a real-world description is the one that gets found.

The bottom line

The portals are not going anywhere, and neither are filters. But a real and growing share of buyers are starting their search by describing their ideal home to an AI, and that share skews younger, which is to say it skews towards the future.

The listings that win in that world are the clear, complete and specific ones. That's always been good practice, and it's now becoming a discoverability advantage too.