Can registered migration agents use ChatGPT or other AI tools?

Yes. The OMARA’s 2026 guidance permits AI use but keeps the agent fully responsible for the accuracy and quality of advice, and all Code of Conduct obligations, including confidentiality and competence, continue to apply.

Do I need client consent to use AI on their matter?

Before entering a client’s personal information into an AI system you need the client’s written consent, unless disclosure is required by law. You should also tell clients in advance that you intend to use AI in their matter.

Is an AI chatbot on a website giving immigration assistance?

It can be. Section 280 of the Migration Act restricts immigration assistance to authorised persons, and the guidance notes that providing it through AI without authorisation may still be an offence.

Is there CPD on AI for migration agents?

Yes. Widen’s self-paced module Artificial Intelligence in Migration Practice is an OMARA-approved Category B activity worth 1 CPD point, covering the guidance, the Code obligations and a practical office policy.

Widen CPD › CPD Guide › AI for migration agents

CPD Guide

AI for migration agents: the OMARA guidance, the Code, and the CPD that covers it

In early 2026 the OMARA published guidance on the use of artificial intelligence in immigration assistance. It does not ban anything. It says the existing rules apply to AI exactly as they apply to a junior staff member or a template: you are responsible for the output, and the client’s information is still confidential.

The guidance in four lines

You may use AI tools, and you remain fully responsible for the accuracy and quality of the advice.
Confidentiality applies. Do not put a client’s personal information into an AI system without their written consent, unless the law requires it.
Tell clients when you intend to use AI in their matter, before you do.
Authorisation still applies. AI cannot give immigration assistance on behalf of someone who is not entitled to give it.

What the guidance actually changes

Nothing in the Code of Conduct was amended for AI. The guidance is the regulator saying out loud how it will read the existing obligations when an agent uses a language model, a document-drafting tool or an AI research assistant. The obligations that matter most:

Code obligation	AI reading
Competence and diligence	An AI-drafted submission is your submission. Citations, visa criteria and policy references must be checked against the current law before anything is lodged. "The tool said so" is not a defence.
Confidentiality	Entering a client’s name, passport details, immigration history or relationship evidence into a third-party AI service is a disclosure to a third person. It needs the client’s written consent, and the service’s terms matter: some retain and train on what you type.
Communication with the client	Clients are entitled to know how their matter is being handled. The guidance expects you to tell them you intend to use AI, in advance, in a way they understand.
Supervision and responsibility	You cannot delegate professional judgement to software any more than to an unregistered employee. Review is not optional.
Section 280 of the Migration Act	Only authorised persons give immigration assistance. An AI chatbot on an unregistered business’s website that answers visa-eligibility questions may be giving immigration assistance unlawfully, with the business liable.

A practical AI policy for a migration practice

Engagement letter clause. State which categories of AI tool you use, for what (drafting, summarising, research), and that personal information is only entered with the client’s written consent. Collect that consent at engagement, not matter by matter.
Tool tiering. Separate tools that are contractually confidential and do not train on your data from consumer chatbots. Client identifiers go only into the first tier, and only when consent exists.
De-identify by default. Most drafting and research tasks work on facts without names. Strip identifiers before the prompt and reinsert afterwards.
Verification step. Every legislative reference, policy instruction and case citation in AI output is checked against the primary source before it leaves the office. Record who checked it.
No AI on the public side. Website chat that answers substantive immigration questions is immigration assistance. Keep public tools to booking and general information, with a registered agent in the loop for anything substantive.
Keep the prompts. If a submission is later queried, the prompt and output history show what was asked, what came back and what you changed.

Where it goes wrong

Hallucinated authorities. Language models produce plausible but non-existent tribunal decisions and legislative references. Courts and tribunals have already sanctioned practitioners in other fields for filing them.
Stale law. Models trained before a change confidently describe the old position. The 2026 regulatory reset for agents is exactly the kind of change a model may not know.
Consent by silence. A privacy policy on a website is not written consent from the client for a specific disclosure.
Staff using personal accounts. The practice’s policy is only as good as the tools your staff actually open at 6 pm.

The CPD that covers it

Widen’s self-paced module Artificial Intelligence in Migration Practice: Responsible Use, Risks and the Code of Conduct is one Category B CPD point built around the guidance and the Code obligations above, with the assessment testing the practical questions: what needs consent, what needs verification, and what must stay out of the prompt. The live CPD Weekend programme includes an AI and ethics workshop as a Category A session when scheduled.

AI in Migration Practice: 1 Category B point, self-paced

Study the OMARA guidance and the Code obligations as they apply to AI tools, pass the 10-question assessment and receive a dated certificate reported to the OMARA. A$30 incl. GST, or one point from a Points Pass. About 90 minutes.

Start the AI module Code of Conduct CPD

FAQ

Can registered migration agents use ChatGPT or other AI tools?: Yes. The OMARA’s 2026 guidance permits AI use but keeps the agent fully responsible for the accuracy and quality of advice, and all Code of Conduct obligations, including confidentiality and competence, continue to apply.
Do I need client consent to use AI on their matter?: Before entering a client’s personal information into an AI system you need the client’s written consent, unless disclosure is required by law. You should also tell clients in advance that you intend to use AI in their matter.
Is an AI chatbot on a website giving immigration assistance?: It can be. Section 280 of the Migration Act restricts immigration assistance to authorised persons, and the guidance notes that providing it through AI without authorisation may still be an offence.
Is there CPD on AI for migration agents?: Yes. Widen’s self-paced module Artificial Intelligence in Migration Practice is an OMARA-approved Category B activity worth 1 CPD point, covering the guidance, the Code obligations and a practical office policy.

← Back to the full CPD requirements guide

General information for registered migration agents, not legal advice. CPD requirements are set by the OMARA under the Migration Agents (CPD Activities, Approval of CPD Providers and CPD Provider Standards) Instrument 2026 (LIN 26/001) — confirm your current obligations and renewal dates at mara.gov.au. Widen CPD is an OMARA-approved CPD provider.

Stay updated

Get an email when a new OMARA-approved CPD module or session goes live, plus occasional CPD reminders. No spam, unsubscribe anytime.

By submitting, you consent to receive CPD update emails from Widen CPD (Dynamic Consultancy Pty Ltd trading as Widen Migration Experts). You can unsubscribe from any email.

Running a migration practice? Widen also does nomination documents, GPS statements, AMSR reports and 407 training plans for agents — see agent services.