Buyers have questions. Your listing needs to answer them.

The search box is becoming a conversation.

Earlier in June, realestate.co.nz launched voice and AI search to find homes. It is in beta, it sits inside the app, and traditional search remains available. But the direction is clear: a buyer can type or speak what they want in plain language and get matched to homes without needing the exact keywords, filters, or listing descriptions.

That matters because it changes the job your listing has to do. Instead of choosing filters for three bedrooms in a suburb, a buyer can describe the life they want: a warm home for two kids near a good school, off the main road, with somewhere to work from home. The portal then has to work out which listings answer that request.

Your listing material now needs to support intent beyond search fields. The description, property facts, photos, floor plan, captions, and first reply all need to make the home easier to understand. If the detail is missing, vague, or unsupported, there is less for a buyer to trust and less for AI-assisted search to work with.

This is bigger than one portal update. On 7 May, OpenAI released new realtime voice models and used Zillow as an example of voice-to-action: a buyer asks for homes inside an affordability range, avoids busy streets, and schedules a tour. That is a US example, not a New Zealand product you can plug into your business today. Treat it as a signal of where interfaces are heading. On 18 June, Google launched Ask Ad Manager, a beta conversational agent that helps publishers ask questions of ad performance instead of reading a dashboard.

Put those signals together and the lesson is not that you need another subscription. It is that your listing has to be easier to understand before the buyer ever reaches you. Buyer attention is fragmented and hard to earn. You reach more of it when your listing is clear enough for a machine to match and useful enough for a person to stop scrolling. That means the heat pump and insulation are named, not implied. The school zone is confirmed, not hinted at. The photos prove the claims. The first reply answers the question the buyer actually asked.

The practical move this week is small: take one current listing and run the readiness check below. It will show you the buyer questions your listing does not yet answer, the proof points you still need, and the photo gaps that make a home harder to match.

Also this week

Fast replies are becoming part of the product. EliseAI and Zillow Rentals highlighted the impact of AI Assist across shared US rental communities from October 2025 to April 2026. Renters who engaged were reported as 43% more likely to apply, 19% more likely to book a tour, and 24% more likely to sign a lease. These are vendor-reported US rental figures, not New Zealand sales evidence. The transferable point is simpler: a fast, relevant first response keeps buyer interest alive.

Your campaign data is starting to answer back. Google introduced Ask Ad Manager on 18 June. It lets publishers ask plain-language questions, troubleshoot line items, generate reports, and navigate settings using their own data. It is not a tool for your listing workflow, and it is in beta. The signal is the interface shift: dashboards are becoming question-answering systems. The habit worth building is sharper questions of your own campaign numbers.

Appraisal proof is becoming part of the pitch. On 22 June, AI PropTech News covered Howsold, a UK agency building an AI pricing tool that pulls live sold prices, local demand, crime data, and flood risk into a pricing view. This is not New Zealand evidence, so the product itself is not the point. The point is that sellers are being trained to test agent pricing with software. If your appraisal depends on confidence more than evidence, that gap will be easier to challenge.

The prompt: AI search readiness check for a listing

The prompt below pressure-tests a listing against the way a buyer might now describe a home. You give it the listing description, confirmed property facts, notes on the photos and floor plan, and who you think the likely buyer is. It then returns buyer-intent searches, missing proof, photo gaps, copy improvements, likely buyer questions, and vendor-report talking points.

We tested it on multiple (synthetic) listing scenarios using ChatGPT and Claude. One example output looked like this:

Buyer-intent searches this listing should match

• "Show me four-bedroom homes in Saint Heliers with a double garage."
• "I want a home with separate living spaces in the eastern bays."
• "Houses with north-facing outdoor areas in Saint Heliers."

Missing facts or proof

• Floor area, land area, and build year.
• Walking time to Saint Heliers beach and village.
• School zone documentation for claimed schools.
• Council rates and title type.
• Number of off-street car spaces beyond the double garage.

Photo or caption gaps

• Add a floor plan showing the room layout.
• Add a photo showing how the indoor living connects to the outdoor entertaining area.
• Show which bedroom could work as a home office.
• Caption the garage as "Double internal-access garage" only if that has been supplied.
• Use "North-facing outdoor entertaining area" only where that exact orientation has been supplied.

That assessment turns a weak listing into an operational checklist. It does not write around missing facts. It tells you what you need to confirm and convey, to make your listing visible.

Paste this into ChatGPT, Claude, Gemini, or Perplexity:

Use the output as an audit, not finished copy. Confirm the facts it flags, add the missing proof, and decide what belongs in the live listing.

The point is not to please a portal. It is to give a buyer, and the systems now helping that buyer, enough clear evidence to understand the home. Start with one current listing. Find the missing facts, add the proof, and tighten the first reply before the next open home.

— The Listing Signal