# AI for New Zealand real estate.

URL: https://levystreet.com/industries/real-estate/
Updated: 2026-10-19

AI for real estate agents in New Zealand: which sales workflows to automate, the REA rules and Fair Trading Act limits, and a first project to start with.

## Short answer

AI for real estate in New Zealand is mostly drafting and sorting: listing descriptions from verified property facts, first replies to buyer enquiries, comparable sales summaries for appraisals and weekly vendor reports. The Real Estate Authority says an error made by an AI tool is no defence to a breach, so a licensee checks every claim, price and disclosure before it reaches a buyer or vendor.

- REA guidance from November 2024 says an AI tool's error is no defence to a licensee's breach.
- Property claims need reasonable grounds when published; Fair Trading Act fines reach $600,000 for agencies.
- Appraisals need written comparable sales evidence; REA says electronic estimates alone are unlikely to be enough.
- Residential property management sits outside the Real Estate Agents Act; tenancy law governs it.
- Listing descriptions from a verified fact sheet make the safest first project.

## Where AI helps a real estate agency

AI helps real estate agents with the writing and sorting around each sale: listing copy, buyer replies, the evidence behind an appraisal, vendor reports and post-offer paperwork. It does not help with the parts buyers and vendors pay for: judgement about price, honest disclosure and negotiation.

Those same parts are where complaints come from. The Real Estate Authority (REA, Te Mana Papawhenua) received 487 formal complaints in the year to 30 June 2025, up 35%, and named customer service, skill and care, disclosure and misleading advertising as the top themes ([REA, 14 November 2025](https://www.rea.govt.nz/about-rea/news/media-releases/rea-reports-increase-in-complaints-in-202425-challenging-market-conditions/)). It reported 15,692 active licences, including 12,300 salespeople. Faster drafts with unsupported claims would add to those complaints, so every workflow here keeps a licensee between draft and customer.

## Real estate sales or property management: which guide fits

This guide covers selling: residential, rural and commercial sales by licensed agents. Rentals have their own section, [AI for property management](/industries/property-management/), because the law is different.

The [Real Estate Agents Act 2008](https://www.legislation.govt.nz/act/public/2008/66/en/latest/) regulates "real estate agency work" done to bring about a transaction, and its definition of transaction excludes a tenancy covered by the Residential Tenancies Act 1986. Managing residential tenancies is therefore not real estate agency work and needs no REA licence; it runs on tenancy law, notice periods and healthy homes records. The Government has proposed a separate registration regime for them ([HUD, last updated 24 March 2026](https://www.hud.govt.nz/our-work/regulation-of-residential-property-managers)); check its status before relying on it.

Where an agency runs both, the two meet: a tenanted listing needs the tenant's permission for viewings. Keep each set of rules in its own workflow.

## Real estate workflow map

The best candidates are text-heavy and easy to check against a source. The worst involve legal documents or regulated decisions.

| Workflow | What AI does | What the licensee decides | Data involved | Difficulty |
|---|---|---|---|---|
| [Listing description drafting](/industries/real-estate/listing-description-drafting/) | Drafts portal, brochure and social copy from a verified fact sheet | Every claim, price wording, vendor sign-off | Fact sheet, title, floor plan, photos | Low to medium: easy to check, costly if wrong |
| [Buyer enquiry handling](/industries/real-estate/buyer-enquiry-handling/) | Matches enquiries to listings, sends approved facts, offers viewing times | Answers on condition and price, every offer | Portal enquiries, listing record, calendars | Medium: live customers, weekend peaks |
| [Appraisal preparation](/industries/real-estate/appraisal-preparation/) | Filters recent sales, drafts the comparables table and reasoning | Which sales count, the price range | Sales data, viewing notes, title | Medium: rule 10.2 evidence |
| Vendor campaign reports | Turns enquiry counts and buyer feedback into a weekly update | Advice on price or campaign changes | CRM activity, feedback notes | Low: numbers come from the CRM |
| Open home follow-up | Drafts follow-up messages and files feedback against the listing | Who receives marketing, and on what consent | Attendee register, notes | Low, if consent is recorded |
| LIM and building report summaries | Lists items from LIMs, building reports and titles with page references | What must be disclosed, and how | LIM, building report, record of title | High: REA warns about relying on AI summaries |
| Sale and purchase agreement checks | Flags blanks and mismatches against the listing record | Conditions, advice to parties, signing | Draft agreement, listing record | High: a legal document |
| AML/CFT due diligence files | Checks that identity documents are present, legible and in date | Risk rating, enhanced checks, reports to Police | Identity documents, company and trust records | High: regulated decisions |

## Which New Zealand rules apply to AI in real estate

No rule bans AI in real estate. Existing obligations apply to whatever a tool produces, and REA's [generative AI guidance](https://www.rea.govt.nz/assets/2024-Uploads/REA-Generative-AI-Guidance-for-Real-Estate-Licensees_November-2024.pdf) (12 November 2024) is blunt: "the fact that any error was made by a Gen AI tool used by the licensee will not be a defence". It adds that supervisors can be held to account for the people they supervise, and encourages every agency to have an AI policy.

| Rule | What it requires | What it means for AI work |
|---|---|---|
| [Code of Conduct](https://www.legislation.govt.nz/secondary-legislation/pco-drafted/2012/413/en/latest/) rules 5.1 and 6.4 | Skill and care; no misleading, false or withheld information | A sent AI draft is the licensee's statement |
| Rules 9.4, 10.4 and 10.9 | Honest price expectations; advertising only on terms the client authorised | Price wording comes from the agency agreement |
| Rules 10.2 and 10.3 | Written appraisal backed by comparable sales | AI assembles evidence; the licensee forms the opinion |
| Rule 10.7 | Disclose known defects; flag likely hidden ones | Report summaries need page references |
| [Fair Trading Act 1986](https://www.legislation.govt.nz/act/public/1986/121/en/latest/) ss 9, 12A, 14 | No misleading claims about land's characteristics, price, location or permitted use; reasonable grounds when made | An unchecked sentence can be unsubstantiated even if true |
| Privacy Act 2020 and rule 9.17 | The information privacy principles (IPPs 1 to 13, including IPP 3A); client confidentiality | Approved tools only; notice for data bought in |
| [AML/CFT Act 2009](https://www.legislation.govt.nz/act/public/2009/35/en/latest/) | Customer due diligence by agents as reporting entities | AI checks documents; people decide risk |
| [Unsolicited Electronic Messages Act 2007](https://www.legislation.govt.nz/act/public/2007/7/en/latest/) | Consent, sender details and an unsubscribe for messages promoting land | AI-written nurture messages inherit all three |

REA's [marketing guidance](https://www.rea.govt.nz/real-estate-professionals/industry-guidance-and-information/listing-a-property/marketing-and-advertising/) puts Fair Trading Act fines at up to $200,000 for licensees and $600,000 for agencies. It does not address AI-edited photos, but REA chief executive Belinda Moffat told [RNZ in July 2026](https://www.rnz.co.nz/life/lifestyle/home/how-far-can-real-estate-agents-go-with-enhancing-photos) that marketing photos should be "a true representation" of the property, however they were enhanced or staged.

Since 1 May 2026, [IPP 3A](https://www.privacy.org.nz/resources-and-learning/a-z-topics/ipp3a/) requires you to tell people when you collect their information from someone else, unless an exception applies; prospecting lists built from owner data are the obvious case. The Department of Internal Affairs became the single AML/CFT supervisor on 1 July 2026 ([Ministry of Justice](https://www.justice.govt.nz/justice-sector-policy/key-initiatives/aml-cft/info-for-businesses/)). This is general information; get legal advice on anything contested.

## The software NZ agencies already run

Most agencies run a sales CRM, a property data subscription and the portals, and several already include AI (as listed on 20 October 2026). MRI Software says its [Vault CRM](https://www.mrisoftware.com/nz/products/vault/) (formerly VaultRE) can "instantly generate property descriptions". Rex Software says [Rex AI](https://www.rexsoftware.com/resources/rex-information-pack) is rolling out through 2026, with AI Admin and AI Prospecting live. Cotality (formerly CoreLogic) sells [Property Guru](https://www.cotality.com/nz/products/property-guru), which supplies sales history, comparative market analysis reports and owner names. Rentals usually sit in a separate system such as MRI Palace.

Listings are now read by AI assistants too: Trade Me Property launched a search app inside ChatGPT in February 2026 ([eCommerceNews](https://ecommercenews.co.nz/story/trade-me-launches-chatgpt-property-search-app-in-nz)), alongside the realestate.co.nz and OneRoof portals. Before switching on a built-in feature, check where it processes data and whether your content trains the vendor's models. Custom work suits drafts grounded in your own records with auditable checks; see our [AI integration](/services/ai-integration/) service.

## A first project: listing descriptions from a verified fact sheet

Listing copy is the safest place to start. Every listing already passes through agent review, vendor sign-off and supervisor oversight, and the main risk, a sentence the facts do not support, is testable.


1. **Design the fact sheet.** One field per claim, each with its source: record of title, LIM, floor plan, or a vendor statement marked verified or not.
2. **Write the banned list.** Subdivision potential, "home and income" and school zones stay out unless a document supports them.
3. **Test on 20 past listings.** Trace every drafted claim back to the sheet.
4. **Run live with review.** The agent edits, the vendor signs off as now, and an automated pass flags sentences with no matching fact.
5. **Measure for a month.** Minutes per listing, unsupported claims caught, vendor change requests.


Illustrative example: an agency lists 30 properties a month (an assumption). Writing the portal, brochure and social copy takes about 60 minutes per listing. If checking a draft takes 20 minutes, the saving is 30 × 40 = 1,200 minutes, about 20 hours a month. At roughly 30,000 input and 3,000 output tokens per listing on Claude Sonnet 5.5 (US$2 and US$10 per million tokens, as listed on [19 October 2026](https://platform.claude.com/docs/en/about-claude/pricing)), model fees are about US$0.09 a listing. Integration and review are the real costs; our [workflow automation](/services/ai-workflow-automation/) service covers the build.

## Where AI goes wrong in real estate


- **Invented features.** A model adds "double glazing" because similar homes have it. Control: copy only from fact sheet fields, checked automatically.
- **Altered images without disclosure.** Virtual staging moves a window, or an edit removes a pylon. Control: label edited images, keep originals, never edit permanent features.
- **Price leakage.** A buyer reply quotes a figure the vendor gave privately. Control: keep vendor price limits out of the model's context.
- **Unverified vendor claims.** REA recommends not publishing claims you cannot verify, such as subdivision potential. Control: unverified fields never reach the copy.
- **Confident report summaries.** REA warns about relying on AI to summarise LIMs and builder reports. Control: page references and a licensee reading the source.


Write these controls into an agency AI policy; our guide to [creating a team AI use policy](/learn/creating-a-team-ai-use-policy/) is a starting point. Levy Street's closest built work is in property operations rather than sales: the [Property Management Agent for APL Property](/portfolio/property-management-agent/), an internal AI agent that handles routine administration (email triage, records, orders, follow-up) for a regional NZ property firm, with staff keeping decisions.

## Questions people ask

### Can AI write real estate listings in New Zealand?

Yes, as a first draft. The Real Estate Authority lists drafting marketing material among the ways licensees already use generative AI, and it says licensees stay responsible for the result. Give the model a verified fact sheet rather than a few notes, ban claims you cannot support, and have the agent check each sentence before the vendor signs off. Unsubstantiated claims can breach the Fair Trading Act even if they turn out to be true.

### Do agents have to disclose virtual staging or AI-edited photos?

We found no REA rule that prescribes a label. The Code of Conduct says a licensee must not mislead, and REA's chief executive told RNZ in July 2026 that marketing photos should be a true representation of the property, however they were edited. A practical standard is to label staged and edited images, keep the originals, and never remove or add permanent features such as power pylons, views or a new roof.

### Can an AI chatbot answer buyer questions about a listing?

It can answer from facts the agent has approved, such as open home times, the address and how to get the information pack. REA's guidance says a non-licensed assistant should not answer enquiries about a property on the seller's behalf. REA has not applied that to AI, but the same caution makes sense, so route questions about condition, boundaries, consents and price negotiation to the licensee.

### Can AI do a market appraisal for a vendor?

It can prepare one. Rule 10.2 of the Code of Conduct requires a written appraisal that reflects current market conditions and is supported by comparable sales. REA says an electronic appraisal or market estimate is unlikely to meet that on its own and expects the licensee to view the property. AI can pull and filter sales and draft the explanation. The licensee chooses the comparables and the range.

### Is it safe to put vendor and buyer details into ChatGPT?

Not into a personal or free account. Rule 9.17 restricts disclosing a client's confidential personal information, and REA's AI guidance warns that providers may see what you enter. Use a business plan or API with training switched off, keep vendors' price limits and personal circumstances out of tools that draft buyer replies, and record where data is processed. If you buy in owner data, IPP 3A now applies to that collection.


## Sources

- [Generative AI Guidance for Real Estate Licensees](https://www.rea.govt.nz/assets/2024-Uploads/REA-Generative-AI-Guidance-for-Real-Estate-Licensees_November-2024.pdf)
- [Real Estate Agents Act 2008](https://www.legislation.govt.nz/act/public/2008/66/en/latest/)
- [Real Estate Agents Act (Professional Conduct and Client Care) Rules 2012](https://www.legislation.govt.nz/secondary-legislation/pco-drafted/2012/413/en/latest/)
- [REA reports increase in complaints in 2024/25 challenging market conditions](https://www.rea.govt.nz/about-rea/news/media-releases/rea-reports-increase-in-complaints-in-202425-challenging-market-conditions/)
- [Marketing and advertising](https://www.rea.govt.nz/real-estate-professionals/industry-guidance-and-information/listing-a-property/marketing-and-advertising/)
- [How far can real estate agents go with enhancing photos?](https://www.rnz.co.nz/life/lifestyle/home/how-far-can-real-estate-agents-go-with-enhancing-photos)
- [Fair Trading Act 1986](https://www.legislation.govt.nz/act/public/1986/121/en/latest/)
- [IPP3A, notification requirements for indirect collection of personal information](https://www.privacy.org.nz/resources-and-learning/a-z-topics/ipp3a/)
- [Anti-Money Laundering and Countering Financing of Terrorism Act 2009](https://www.legislation.govt.nz/act/public/2009/35/en/latest/)
- [AML/CFT information for businesses](https://www.justice.govt.nz/justice-sector-policy/key-initiatives/aml-cft/info-for-businesses/)
- [Unsolicited Electronic Messages Act 2007](https://www.legislation.govt.nz/act/public/2007/7/en/latest/)
- [Regulation of residential property managers](https://www.hud.govt.nz/our-work/regulation-of-residential-property-managers)
- [MRI Vault (formerly VaultRE)](https://www.mrisoftware.com/nz/products/vault/)
- [Rex information pack](https://www.rexsoftware.com/resources/rex-information-pack)
- [Property Guru](https://www.cotality.com/nz/products/property-guru)
- [Trade Me launches ChatGPT property search app in NZ](https://ecommercenews.co.nz/story/trade-me-launches-chatgpt-property-search-app-in-nz)
- [Claude API pricing](https://platform.claude.com/docs/en/about-claude/pricing)

Published by Levy Street (https://levystreet.com), Wellington, New Zealand.