The short answer
AI cold calling in Australian real estate has no independently published evidence base, so nobody can honestly say it works yet. The figures that circulate come from vendor pages that state no sample, no period and no definition of the unit counted. Judge it instead on connected calls, conversations held, appraisals booked and listings won.
What to take away
- We have found no independently published Australian case study with real numbers on AI voice calling in real estate, and the vendor figures that do circulate state no sample, no period and no definition.
- Judge any vendor on four units you can audit yourself: connected calls, conversation length, appraisals booked and listings won. Everything else is derived, and derived in the vendor's favour.
- The human benchmark is public. About 12 dials per appraisal on an owned Australian list, 330 calls per appointment on a cold United States list, and a 5.4 per cent connect rate on cross industry cold dials.
- Australian buyers are the most sceptical of the four markets Cotality surveyed, and 68 per cent want to be told when AI is involved, so disclosure is a commercial decision before it is anything else.
What would it even mean for AI cold calling to work?
Nothing, until you say which unit you mean. A dial, a connected call, a conversation, an appraisal booked and a listing won are five different things, and a vendor can be perfectly truthful about one of them while you are buying another.
Start with the honest part. As at August 2026 we have not found an independently published Australian case study putting real numbers on AI voice calling in residential real estate. We looked: industry press, the regulators, the proptech surveys, and the vendors themselves. Figures do circulate, on Australian vendor pages as well as overseas ones, and not one of them publishes a sample size, a date range or a definition of the unit being counted. That is the most useful thing a buyer can be told, because it means every claim you are shown about AI cold calling is either a vendor's own arithmetic or a benchmark borrowed from human calling. So the question worth asking is not whether it works. It is what you would measure to find out.
None
Independently published Australian case studies on AI voice calling in residential real estate that state a sample, a period and a definition of the unit counted, as at August 2026, after a search of industry press, regulator publications, proptech surveys and vendor sites. Vendor pages carrying figures without those three things are not counted, and there are several.
ListingLine desk review, August 2026
| Unit | What it means | Nearest published benchmark | Source |
|---|---|---|---|
| Dial | One attempt placed | 141,131 dials in a single 12 hour phone day | Elite Agent, February 2026 |
| Connected call | A person answers and speaks | 5.4 per cent of cold dials connect for an average rep | Gong Labs, July 2024 |
| Conversation | The call gets past the opener | Successful cold calls run about 5 minutes 50 seconds | Gong Labs, 2021, updated March 2026 |
| Appraisal booked | A time in the diary | About 12 dials per appraisal on an owned list | Elite Agent, Ray White phone days 2026 |
| Listing won | An agency agreement signed | 6.5 per cent of phone sourced appraisals became listings | Dan White, Ray White, calendar 2025 |
Notice which column is doing the work. Every benchmark in it was produced by humans on the phone. Nobody publishes an Australian connect rate for AI calls, ours included, because an honest version of that number needs a stated sample, a stated period and a definition of what counts as connected. Until a vendor gives you all three, treat the figure as marketing.
What does the human benchmark look like?
Human phone work is the only side of this with published numbers behind it, so it sets the bar an AI caller has to clear. Ray White publishes the raw counts from its twice yearly network phone days: 141,131 calls and 12,304 appraisals in February 2026, which is 11.5 dials per appraisal booked, on lists the agents already owned.
| Measure | Owned Australian list | Cross industry cold dials | Cold United States property list |
|---|---|---|---|
| Answered or connected | Not published | 5.4 per cent average, 13.3 per cent top quartile | 28 per cent answered |
| Dials per meeting or appraisal booked | 11.5 to 12.7 | 402 average, 45 top quartile | 330 |
| What happens next | 6.5 per cent of appraisals became listings | Set rate 4.6 per cent average, 16.7 per cent top quartile | 19 appointments and 11 referrals from 6,264 calls |
Two things fall out of that table. Cold lists are an order of magnitude worse than owned lists, and the cross industry connect rate of 5.4 per cent means an average human rep speaks to about one person in every nineteen dials, on Gong Labs analysis of more than 300 million cold calls. An AI caller that beats nineteen dials per conversation is doing something real. One that does not is an expensive way to leave voicemails.
What is actually measurable today?
Four things, and all four sit on your side of the transaction. Connected calls, because your telephony records count them. Conversation length, because every recording has a duration. Appraisals booked, because they land in your diary. Listings won, because you sign them. Everything else a vendor reports is derived, and derived numbers are where the flattering definitions live.
The trap is the denominator. A vendor quoting a connect rate can choose dials, unique numbers, or numbers that actually rang. A vendor quoting an appraisal rate can count a diary entry, a confirmed appointment, or an appraisal that happened. Ask which, in writing, before the first campaign runs.
What should you demand from a vendor before you pay?
Six answers, in writing, before any money moves. None of them is unreasonable, and a vendor who cannot answer them has told you something useful anyway.
Six questions to put to an AI calling vendor
Define the billing unit
Ask exactly what triggers a charge: a dial, a ring, a connected call, a minute of speech, or a booked outcome. Then ask what happens when an answering machine picks up, because that single answer can double or halve your real cost.
Ask for the aggregate, not the anecdote
Ask for outcomes across all accounts over a stated period with the sample size attached. One agency's good month is not evidence, and a vendor with no aggregate to show you has not been measuring the thing you are buying.
Read the opening words
Ask for the exact script the agent opens with, including whether it says it is artificial and how it answers the question are you a robot. Read those words aloud in your own office before you approve them.
Check the washing and the calling hours
Ask how the list is washed against the Do Not Call Register, how often, and who performed the wash. The Do Not Call Register Act 2006 defines a voice call to include a call made with a recorded or synthetic voice, so a synthetic caller sits inside the same definition as a human one. Our read of those obligations is in the compliance overview, and none of it is legal advice.
Ask what happens to the recordings
Ask where the audio is stored, for how long, who inside the vendor can listen to it, and what the homeowner is told at the start of the call. Recording obligations in Australia are set state by state, so have your own adviser look at the answer.
Run a paid pilot on your own records
Insist on a small paid pilot scored against your telephony records and your diary rather than the vendor dashboard. Agree the success threshold before it starts, expressed in appraisals booked, not in calls made.
What would make it not work?
Three failure modes, and the first one is not technical. Australians are the most sceptical buyers of the four markets Cotality surveyed: only 40 per cent would accept an AI valuation of a home, against 62 per cent in Canada, 52 per cent in the United States and 46 per cent in the United Kingdom. A homeowner who feels handled does not book an appraisal, and they remember which agency did it.
68 per cent
of homebuyers want clear notification when AI generates a listing, a price or a mortgage recommendation. Only 40 per cent of Australian buyers would accept an AI valuation, the lowest of the four markets surveyed.
Cotality, AI in Housing 2026, fieldwork 29 January to 9 February 2026
The second is disclosure. In May 2026 the ARC Centre of Excellence for Automated Decision Making and Society documented an AI agent cold calling Australian property owners under the name Daniel. Its researchers noted the system did not initially disclose that it was an AI agent, or that the call was being recorded.
It's supercharging something that was already happening with endless cold calls.
That matters commercially before it matters anywhere else. Cotality found 37 per cent of buyers want AI disclosure made mandatory, rising to 61 per cent of Baby Boomers and falling to 25 per cent of Gen Z. Baby Boomers are the people who own the houses in most farm areas.
The third is the list. An AI caller multiplies whatever list you point it at, errors included: wrong owner, sold last year, deceased, or a number on the Do Not Call Register that nobody washed. Ten times the dials is ten times the exposure, and the obligation sits with whoever caused the call to be made.
Where does that leave a principal deciding this year?
With a category that is plausible and unproven, which is uncomfortable but accurate. The mechanism is sound: dialling is the part of the job that does not need a licensed human, and the arithmetic of calls per listing says the dials set the ceiling on everything else. What is missing is published evidence that a machine holds a conversation an Australian homeowner acts on.
ListingLine is an AI voice agent that calls an Australian real estate agent's farm area in the agent's name, at A$2.00 per connected call. We have not published an aggregate outcome across accounts, so by the standard set out above you should not treat anything we say about results as proof. Put the six questions to us as well.
The right test is small, paid, and scored on your own records. If the words are what you care about, the appraisal call script is what a good version of this call sounds like, and the rules the dialling has to stay inside do not soften because the caller is synthetic.
Common questions
- Has anyone published Australian results for AI calling in real estate?
- Not that we could find as at August 2026. We searched industry press, regulator publications, the trans Tasman proptech surveys and the vendors themselves. Everything carrying a number is United States vendor marketing with no sample size and no date of data attached. When a credible Australian dataset appears we will link it here, including if the numbers are unflattering to this category.
- How is a connected call different from a dial?
- A dial is an attempt. A connected call is a dial where a person answers and speaks. The gap between the two is enormous: Gong Labs, analysing more than 300 million cold calls in July 2024, found an average sales rep connects on 5.4 per cent of dials and a top quartile rep on 13.3 per cent. Pricing quoted per dial and pricing quoted per connected call are not comparable numbers.
- Do homeowners hang up once they realise it is an AI?
- We have found no published Australian measurement of that, so anyone quoting you a hangup rate is quoting themselves. What is measured is the sentiment around it. Cotality's four country buyer survey, fielded in early 2026, found 68 per cent want clear notification when AI is involved, and only 40 per cent of Australian buyers would accept an AI valuation, the lowest share of the four markets.
- Does an AI caller sit inside the same calling rules as a person?
- The Do Not Call Register Act 2006 defines a voice call to include a call made with a recorded or synthetic voice, which puts a synthetic caller inside the same definition as a human one. Washing, permitted hours, identification and termination obligations follow from that definition. None of this is legal advice, and an agency should have its licensee in charge sign off the process before any campaign runs.
- What does a fair pilot look like?
- Paid, small, and scored on your telephony records and your diary rather than a vendor dashboard. Agree the billing unit before it starts, set the threshold in appraisals booked rather than calls made, run it on a list you washed yourself, and listen to at least twenty recordings end to end. A fortnight is usually enough to see whether real conversations are happening.
- Why publish a piece that does not conclude your own product works?
- Because the alternative is a page an agent stops believing halfway down. We sell calls, so anything we say about our own connect rates is an interested party talking. The useful thing we can offer is the measurement framework and the published human benchmarks, both of which count against us when the calls are poor. That is exactly why they are worth printing.
Sources
- 1Ray White network books 12,000 appraisals in 12 hour phone blitz. Elite Agent, Event 26 February 2026.
- 2Ray White members hit the phones for fast paced appraisal drive. Elite Agent, Event June 2026.
- 3The hidden power of cold calling: insights from 300M calls. Gong Labs, Published 18 July 2024, dataset of more than 300 million calls.
- 4Has Cold Calling Gone Cold?. Keller Center for Research, Baylor University, Calls placed November 2011, published September 2012.
- 5Trust but verify: what buyers expect from AI in housing. Cotality, AI in Housing 2026, four country buyer survey, Fieldwork 29 January to 9 February 2026.
- 6Australia lags global peers on property AI adoption. Cotality, Published 26 May 2026.
- 7AI cold calling real estate assistant raises issues of consent, trust and transparency. ARC Centre of Excellence for Automated Decision Making and Society, Published 29 May 2026.
- 8Do Not Call Register Act 2006, section 4 definition of voice call. Federal Register of Legislation, Compilation in force 1 September 2021.
Written by
Jason
Founder, ListingLine
Builds the AI voice agent that ListingLine customers point at their farm area. Spends most of his week reading call transcripts, which is an unusual and fairly grim way to learn what Australian homeowners will and will not talk about on the phone.