Software that predicts who will sell: what privacy law allows

In this article

Written by

Kooky

Builder of Shaka, the payment router that pays every agent their commission on closing date.

About Kooky and Shaka →

A letter arrives from a local agency a few months before the owners had told anyone they were thinking of selling. It may be coincidence. It may also be arithmetic: a growing class of prospecting software ranks properties, and the people who own them, by how likely they are to come to market soon. The agency sees a list sorted from most likely to least, and starts at the top.

Nothing in Australian law bans that kind of prediction. What the law does is attach conditions to each stage of it: where the information came from, whether the owner was told, what the score is used for, how accurate it is, and what happens when the owner asks to see it or to be left alone. This guide works through those conditions for a Queensland agency principal weighing up such a tool, and for a homeowner who suspects they have been scored. It deals with the Commonwealth Privacy Act 1988 and the guidance of the Office of the Australian Information Commissioner (OAIC), with a short stop at Queensland's own conduct standards for agents. What follows is the general rule, not a ruling on any one case.

OAIC small business page, OAIC Australian Privacy Principles Guidelines chapters 1 and 7.

How a seller score is built

Vendors publish little about the inside of their models, so the description here is limited to what one large data company says on its own product pages. The property data company Cotality describes a product called Propensity to List as a model that identifies properties more likely to be listed for sale in the next three months. Its page says the model draws on searches and activity on the company's platforms, on property transfers and listings, and on market trends. The page presents that product for lenders that want to reach customers before they refinance elsewhere, and it claims the model delivers a customer segment 15 times more accurate than a random sample, a figure a footnote attributes to the company's internal testing.

The same company's page for its RP Data service lists a prospecting filter for tenure, withdrawn listings and aged listings, described as a way to find properties held for about the average holding period in an area. It also lists marketing contacts, described as a database of names and phone numbers for direct marketing.

Those two pages are enough to show the general shape. A seller score combines facts about a property (how long it has been held, whether it was listed and withdrawn), signals of interest (searches, activity on a platform) and market conditions, then expresses the result as a likelihood. The output is a prediction about what a household will do.

Is the agency covered by the Privacy Act?

The first question is whether the Act applies to the agency at all. The rules sit in sections 6C to 6EA of the Act, and the OAIC's small business page states the starting point: a small business is one with an annual turnover of $3 million or less, and most small businesses are not covered.

The same page then lists small businesses that are covered whatever their turnover. Several matter to a real estate office.

Two rows deserve a closer look. The first is trading in personal information. The OAIC's page on the subject says a business trades when it collects or discloses someone's personal information for a benefit, service or advantage, and it gives buying a mailing list without the consent of everyone on it as an example. An agency that buys a contact list to feed a prospecting tool may therefore fall under the Act through its own conduct, however small it is.

The second is anti-money laundering law. The OAIC's guidance for reporting entities, published on 27 February 2026 and updated on 28 August 2026, records that the Privacy Act applies to real estate professionals from 1 July 2026, once they become reporting entities. The coverage is partial. The guidance says small businesses are covered for activities connected with their obligations under the anti-money laundering legislation, and not for their other business activities unless another ground applies. It gives the case of an agency with $1.1 million in turnover: covered for the customer and personnel due diligence it carries out, not covered for the rest. Prospecting is not due diligence, so this route does not, on its own, bring a small agency's seller scoring under the Act.

The proposal to end the small business exemption

The exemption has been under review for years. The Attorney-General's Department's page on the Privacy Act Review Report says the report's proposals would address the risks posed by the small business, employee records, political and journalism exemptions in a proportionate and practical way. The Government's response, released on 28 September 2023, listed small business privacy reforms among the subjects of further targeted consultation, alongside a detailed analysis of costs and benefits.

The Privacy and Other Legislation Amendment Act 2024, which the Parliament of Australia's bill record shows received assent on 10 December 2024 as Act No. 128 of 2024, left the test where it was: section 6D of the Act, in the compilation dated 4 June 2026, still sets it at $3,000,000. On 31 August 2026 the department opened consultation on an exposure draft, the Privacy Amendment (Personal Data Protection) Bill 2026, which closed on 18 September 2026. The consultation paper describes a package of roughly 40 proposals. Neither the paper nor the draft Bill proposes removing the exemption or changing the turnover test. The draft does rework one of the exceptions: it would add a definition of trade, covering a disclosure of personal information for money or other consideration or for the purposes of direct marketing, and rewrite the trading paragraphs of section 6D to match. The department's consultation page states that the Bill remains subject to further consideration by government.

So the position on 9 October 2026 is unchanged in law: the $3 million test and its exceptions decide coverage. A smaller agency may opt in, a mechanism the OAIC's page describes. The rest of this guide assumes an agency that is covered, by size, by conduct or by choice.

A score about a person is personal information

The OAIC does not treat a prediction as the business's own work product. Its guidance on privacy and the use of commercially available AI products, published on 21 October 2024, says inferred, incorrect or artificially generated information produced by AI models is personal information where it relates to an identified or reasonably identifiable individual, and must be handled under the privacy principles. Its companion guidance on developing and training generative AI models, published the same day, uses the same wording.

The guidance goes one step further. When a system generates or infers personal information, that generation counts as a collection. Chapter 3 of the OAIC's privacy principles guidelines, updated on 13 May 2026, says the same about information created from other information an entity already holds, and names AI, automated decision making and data analytics as examples.

Whether a given score is "about" an individual depends on the facts. A ranking of streets by turnover is market analysis. A flag against a named owner's record that says "likely to sell within three months" says something about that person's intentions and circumstances.

Collection: necessity, fair means and notice

Once the score is a collection, Australian Privacy Principle 3 sets three tests, each explained in chapter 3 of the guidelines.

The information must be reasonably necessary for the agency's functions or activities. The guidelines call this an objective test and say collection that is merely helpful, desirable or convenient does not pass, nor does collecting information in case it proves useful later.

The collection must be by lawful and fair means. The guidelines say it would usually be unfair to collect personal information covertly, without the individual's knowledge, and that giving notice does not by itself make a collection fair. They also deal with public records: the fact that information is publicly available does not allow it to be collected and used in whatever way a business chooses.

The information must be collected from the individual unless that is unreasonable or impracticable. A seller score is never collected from the owner, so an agency relies on this exception every time. The guidelines list the factors: what the individual would reasonably expect, how sensitive the information is, the privacy risk of the other source, and cost. They add that a business cannot lean on the exception to buy data from a broker simply because asking people directly is inconvenient.

Principle 5 then asks the agency to take reasonable steps to tell the individual about the collection, at or before the time or as soon as practicable afterwards. Chapter 5 of the guidelines says that where information comes from a third party, or the individual may not be aware of the collection, the notice should cover the fact and circumstances of collection, including the source, or the kinds of sources if naming one is impracticable. For a household that has never dealt with the agency, the practical question is when, and how, the owner finds out. The guidelines accept that sometimes no step is reasonable, but the business must be able to justify that.

Using the score: secondary use and direct marketing

Two principles govern use, and which one applies depends on what the agency does with the information.

Principle 6 covers use and disclosure in general. Chapter 6 of the guidelines explains that information may be used for the primary purpose it was collected for, and for a secondary purpose with consent, where the individual would reasonably expect it and the purpose is related to the first, or under one of the other listed exceptions. That matters most for the agency's own records. Details given by a tenant applying for a rental, a buyer registering at an open home or a past client were collected for those dealings. Loading them into a model, or handing the database to a software supplier, is a further use or possibly a disclosure, and the OAIC's AI guidance says entering personal information into an AI system is governed by Principle 6 either way. It adds that where reasonable expectations cannot be clearly established, a business should seek consent or offer a meaningful opt-out.

Principle 7 takes over when the purpose is direct marketing, which chapter 7 of the guidelines defines as using or disclosing personal information to communicate directly with an individual to promote goods and services. Choosing which owners receive an appraisal offer is that. The chapter separates two cases.

Where the agency collected the details from the individual and that person would reasonably expect marketing, it may proceed if it offers a simple way to opt out. Where the details came from someone else, or the individual would not expect the use, the agency needs consent unless obtaining it is impracticable, and every communication must carry a prominent statement that the person can opt out. Most owners on a purchased or modelled list fall into the second case.

Marketing that does not use personal information to select its recipients sits outside the principle: the guidelines give unaddressed mail to "the householder" as the example. A leaflet run to every letterbox in a suburb is one thing. A run to the addresses a model ranked highest is a selection made with data, and the analysis changes if that data identifies the owners.

Quality: a prediction is a kind of opinion

Principle 10 requires reasonable steps to ensure personal information collected is accurate, up to date and complete, and that information used is also relevant to the purpose. A likelihood cannot be accurate in the way a settlement date can, and chapter 10 of the guidelines offers the nearest guide in its treatment of opinions. An opinion is not inaccurate merely because the individual disagrees with it. It may be accurate if it is presented as an opinion and not as fact, and if it is an informed assessment that takes competing facts into account.

Applied to a seller score, that suggests a record that says what it is: a model's estimate, produced on a given date, from given kinds of input. The OAIC's AI guidance warns that AI systems are known to produce inaccurate results and points to human oversight. For third-party data, chapter 10 suggests checking the supplier's quality practices through contract terms or due diligence before collecting.

A seller score is a guess about a household. The law treats the guess as that household's information.

What changes on 10 December 2026

The 2024 amending Act added three paragraphs to Principle 1, on automated decisions. Its commencement table starts them 24 months after assent, and chapter 1 of the OAIC guidelines, reissued as version 2.0 on 30 September 2026, gives the date as 10 December 2026.

From then, a covered business must add information to its privacy policy if three things are true: it has arranged for a computer program to make a decision, or to do something substantially and directly related to making one; the decision could reasonably be expected to significantly affect an individual's rights or interests; and the individual's personal information is used in the program's operation. The policy must then describe the kinds of personal information used and the kinds of decisions involved.

Whether seller scoring reaches that threshold is open: the chapter gives no example drawn from real estate prospecting. It says the effect must be more than trivial and gives examples such as insurance approvals, access to significant services, and housing among the areas where interests are significant. It says that using programs to target individuals with content and advertisements may have a significant effect, and the examples it attaches are personalised pricing for significant goods or services and limits on access to employment opportunities. A decision to send someone an appraisal letter looks a long way from those. The three conditions are tested tool by tool, so the answer for one that ranks rental applicants may differ from the answer for prospecting.

The statutory tort, the Spam Act and the Do Not Call Register

The same Act created a right to sue for serious invasions of privacy, in a new Schedule 2 to the Privacy Act. Its commencement table set the start at a day fixed by proclamation or, failing that, the day after six months from assent, and the OAIC's page on the tort records that it commenced on 10 June 2025. The schedule applies to conduct, not to a class of regulated business, so the turnover test is not part of it.

The bar is high. A plaintiff must show an intrusion upon seclusion or a misuse of information about them, a reasonable expectation of privacy, an invasion that was intentional or reckless and serious, and a public interest in privacy that outweighs any competing public interest. Proof of damage is not required. Damages for non-economic loss, together with any exemplary damages, are capped at the greater of $478,550 or the maximum available in defamation proceedings. Ordinary prospecting from property records is hard to fit within those elements.

The channel rules sit beside all of this and were covered in the magazine's guide to agent prospecting. In brief, the Australian Communications and Media Authority states three requirements for commercial emails and texts: consent, accurate identification of the sender, and a working unsubscribe that is honoured within five working days. The Do Not Call Register's industry page says a call with any commercial purpose cannot be made to a registered number, and that both the caller and the business behind the call must comply. Chapter 7 of the OAIC guidelines notes that Principle 7 gives way to those two Acts where they apply, and still governs what they exempt. A high score is not consent under any of them.

Queensland's conduct standards still apply

Queensland adds no privacy rule of its own for private agencies, but the Property Occupations Regulation 2014 sets conduct standards for licensees and salespersons in Part 5, made under section 235 of the Property Occupations Act 2014. Two bear on what happens after a model produces a name.

Section 19 requires an agent to take reasonable steps to find out or verify the ownership and description of a property before listing or auctioning it. A data product's owner field is a lead, not that verification. Section 21 requires an agent to find out whether another agent already holds an appointment, and not to solicit or accept one where the client could become liable for a second commission or for damages unless a written statement is given first. A filter built on withdrawn and aged listings will, by design, surface owners who recently had an agent, so that check matters for exactly the leads such a tool produces.

What an owner can ask an agency

An owner who wants to know why an agency wrote, called or knocked has several questions available, and the answers depend on whether the agency is covered by the Act.

• Is the agency bound by the Privacy Act? A covered business must have a privacy policy, and from 10 December 2026 that policy may need to describe automated decisions.
• Where did the agency get the details? Under Principle 7, an individual may ask an organisation for the source of the information used to market to them. Chapter 7 of the guidelines says the organisation must answer unless that is unreasonable or impracticable, generally within 30 days and without charge.
• Stop marketing, and stop passing details on. The same principle lets an individual ask not to receive direct marketing and ask that their information not be disclosed to help other organisations market. The guidelines put a reasonable period for stopping at generally no more than 30 days.
• What does the agency hold? Principle 12 gives a right of access on request, with no form or reason required. Chapter 12 puts a reasonable period for an organisation at no more than 30 calendar days. An organisation may charge for giving access if the charge is not excessive, but not for the request.
• Is it wrong? Principle 13 requires reasonable steps to correct information that is inaccurate, out of date, incomplete, irrelevant or misleading, free of charge. If the agency declines, the owner may ask that a statement of their view be attached to the record.

One limit applies to access. Chapter 12 lists, among the grounds on which an organisation may refuse, that access would reveal evaluative information generated in a commercially sensitive decision-making process, and gives a scorecard's weighting system as an example. The ground covers the evaluative material only, not the personal information the assessment was built on, and the written refusal may include an explanation. An owner may therefore be entitled to see the inputs held about them even where the workings of the model stay closed.

Where a covered agency refuses or does not respond, chapter 12 describes the path: a written complaint to the business first, about 30 days for a reply, then a complaint to the Information Commissioner. Where the agency is not covered by the Act, those rights are not enforceable against it, though the spam and telemarketing rules still apply to how it makes contact.

Kooky, from Shaka

Kooky edits Queensland Estate and builds Shaka, the payment router he made for Queensland property professionals. One payment comes in, and every agent, agency and party in the deal receives their signed share on closing date.

About Kooky and Shaka →