AI for conveyancing: what to automate and what to keep

By Michael Nadalin, Founder, Market Lead · 5 October 2026 · 8 min read

It is 4.40 pm on a Friday. A purchaser client has emailed a 96 page contract for sale with a strata report attached, the agent wants it signed before the weekend, and you have three settlements booked for Monday. The fee is fixed and it is not large. This is the moment AI vendors say they were built for, and for once they are partly right.

Conveyancing is high volume, document heavy and price competitive, so it suits automation better than most practice areas. It is also the area where a single missed easement, a wrong settlement figure or a redirected payment turns into a claim. The useful question is not whether to use AI. It is which tasks you hand over, which you supervise closely, and which you never delegate.

What follows is a working division of labour for an Australian property practice, written on the assumption that you remain personally responsible for everything that leaves your office, because you do.

Start with where the time goes in a file

Before buying anything, map a standard purchase file from instruction to registration and mark each step as reading, writing, checking, deciding or certifying. Most practices find that reading and writing fill the day, while the risk concentrates in checking, deciding and certifying.

That distinction matters because current AI tools are good at reading and writing, uneven at checking, and not permitted to decide or certify. A tool that drafts your initial advice letter in two minutes saves real time. A tool that claims to settle the matter for you is describing something the electronic conveyancing rules do not allow.

What AI does well in a property practice

The strongest use is first pass review of long, structured documents: the contract for sale and its annexures in New South Wales, the section 32 vendor statement in Victoria, the seller disclosure documents in Queensland, strata and owners corporation records, and leases in a commercial acquisition. Ask the tool to extract special conditions, list every departure from the standard form, and identify each easement, covenant and encumbrance with a page reference. Then check its work against the document.

The second strong use is translation into plain English. Clients do not read a fourteen page advice letter. A clear summary of what they are buying, what the risks are and what dates they must meet is something AI drafts well from your own notes, and it is usually the part of the job that gets squeezed on a fixed fee.

The third is the administrative work around the file: chronologies, critical date lists, routine update emails and requisition responses built from your own precedents. None of it is glamorous. All of it adds up.

Where it goes wrong

The failures in property work are quieter than the invented case citations that make the news. A model reviewing a contract will rarely fabricate a clause. It is more likely to miss one, particularly in a scanned annexure with poor image quality, a handwritten amendment, or a plan where the easement is shown graphically and described nowhere in the text. Silence in an AI summary is not evidence that nothing is there.

The second failure is jurisdiction. Property law in Australia is state and territory law, and the differences are not cosmetic: disclosure regimes, cooling off rights, duty, and standard contract terms all vary. A general purpose chatbot will happily blend Victorian and New South Wales concepts into one confident paragraph. It may also describe a regime as it stood before a recent change. Queensland's seller disclosure regime under the Property Law Act 2023 is a good example of something a model with older training data can get wrong.

The third is confidentiality. A contract contains names, addresses, prices and often financial details. Pasting it into a consumer tool that retains or trains on inputs is hard to reconcile with your duty of confidentiality, regardless of how convenient it was.

What must stay with a person

Some steps are not automatable as a matter of law, whatever the technology can do. Under the Electronic Conveyancing National Law and the participation rules made under it, the subscriber verifies identity, establishes the client's right to deal, holds a proper client authorisation, retains supporting evidence and gives certifications when signing. Those are personal obligations with compliance examinations behind them.

Others stay human because of risk. Bank account details and payment directions are confirmed by a person, by telephone, on a number obtained independently of the email that supplied the details. Tax and withholding questions at settlement, including foreign resident capital gains withholding and GST withholding, need a person to confirm the current position and the documents held. So does any advice that asks the client to accept a risk: an unapproved structure, a short settlement, a problematic special condition. The client is entitled to your judgment on those, and your insurer expects it.

Introducing it without betting the practice

Start small and start with something you can verify in under a minute. Client updates and date lists are the obvious first step, because an error is visible at a glance and costs little. Move to document review only once you have a written checking procedure: what the reviewer confirms against the source, and how that is recorded on the file.

Run the tool alongside your existing process for a period before relying on it. Give it contracts you have already reviewed and compare the results. You will learn quickly what it misses, and that list becomes your checklist. Tell your staff plainly which tools are approved for client documents and which are not, and consider whether your cost disclosure and engagement terms should say anything about how you use the technology.

The test to apply to any tool

Ask three questions of any product pitched to a property practice. Does it show you where in the source document each statement comes from, so you can check it? Does it know which Australian jurisdiction it is working in and say so? Does it keep your client's documents confidential on terms you have read? A tool that fails any of the three will cost you more in checking and worry than it saves.

For the research side of property work, such as an unusual covenant, a caveat dispute or a question about a disclosure obligation, the same test applies. A tool like Legal Brain is built around Australian sources and citations you can verify, which is the minimum standard. You still read the authority before relying on it.

The practices that do well with AI in conveyancing will not be the ones that automate the most. They will be the ones that are clear about the line: machines read and draft, people check, decide and certify. Fixed fee work rewards that clarity, and so does your insurer.

Frequently asked questions

Can AI review a contract for sale of land in Australia?

It can do a useful first pass: extracting special conditions, flagging departures from the standard form and listing encumbrances with page references. It cannot replace your review. Scanned plans, handwritten amendments and poor quality annexures are commonly missed, so the title, plans and any flagged items still need to be read by a lawyer or licensed conveyancer.

Is it safe to upload client contracts to ChatGPT or similar tools?

Not without checking the terms first. Contracts contain personal and financial information covered by your duty of confidentiality, and some consumer tools retain inputs or use them for training. Use a tool your firm has approved after reading its terms on retention, training and where data is stored.

Can AI handle settlement on PEXA for me?

No. Under the electronic conveyancing framework the subscriber is responsible for verifying identity, establishing the right to deal, holding a client authorisation and giving certifications when signing. Software can prepare information for you, but the signing and certifying are personal obligations.

Will AI replace conveyancers and property solicitors?

It will replace some of the reading and drafting time on each file, which matters in fixed fee work. It does not replace the steps that carry the legal responsibility: identity, certification, money handling and advice on risk. The more likely change is that practices using it well can handle more files at the same standard.