Artificial intelligence tools are increasingly utilised to explain unfamiliar concepts, summarise documents and produce apparently professional legal submissions within seconds.
For a person facing legal costs, it may be tempting to use artificial intelligence instead of engaging a lawyer. However, there is a significant difference between using AI as an administrative tool and relying on it to identify your legal rights, develop your case and prepare documents for court.
Representing yourself with AI-generated material may appear to save money initially. If the technology misunderstands the law, overlooks a deadline or prepares defective court documents, however, the consequences can be far more expensive than obtaining legal advice at the outset.
AI Can Write Like a Lawyer Without Thinking Like One
Generative AI can produce polished and authoritative-sounding text. That does not mean it understands the law or whether its answer is correct.
Tools such as ChatGPT generate responses by predicting likely combinations of words from the information on which they were trained. They do not reason through a legal problem in the same way as a qualified lawyer and have no independent understanding of truth.
This distinction is particularly important in legal proceedings. The outcome of a case may depend on matters such as:
– the precise cause of action or ground of review available;
– whether the correct court or tribunal has been approached;
– strict limitation periods and procedural deadlines;
– the evidence required to prove each element of the claim;
– whether particular evidence is admissible;
– the remedies the court has power to grant; and
– legislation and court decisions specific to the relevant jurisdiction.
An AI chatbot may produce a document that looks impressive while overlooking the issue that ultimately determines whether the case succeeds or fails.
The Queensland Courts’ guidelines for non-lawyers expressly state that generative AI is not a substitute for a qualified lawyer, cannot provide tailored legal advice and should not be used as a person’s sole or main source of legal information.
Recent Warnings for Self-Represented Litigants
In Asif v Minister for Immigration [2026] FedCFamC2G 1402, a self-represented applicant used ChatGPT to assist with written submissions challenging the refusal of a student visa.
The submissions were filed late and were described by the Court as unnecessarily lengthy, frequently repetitive and containing little substantive legal argument. When the applicant appeared at the hearing, he was unable to explain or expand upon the submissions or make a coherent argument about the legal error he alleged had occurred. The Court found that he did not understand what was contained in the documents filed on his behalf.
Judge Humphreys observed that AI-generated submissions create additional work for courts and opposing parties because they must examine lengthy material to determine whether it contains any genuine legal merit. His Honour warned that stronger procedural responses may be required where submissions have clearly been generated by AI or filed contrary to court orders.
The application was dismissed and the applicant was ordered to pay the Minister’s costs. Because the late AI-generated submissions required additional work from the Minister, the Court also considered whether costs above the usual scale should be ordered.
A similar warning was given in Ba v Sterling Parts Australia Pty Ltd [2026] FedCFamC2G 1245. In that case, a self-represented applicant brought proceedings alleging that he had been incorrectly treated as an independent contractor and unlawfully dismissed after raising concerns about his working arrangements and workplace entitlements.
The applicant filed an evidence list referring to documents said to support his claim. However, when ordered to produce those documents, he failed to do so, apart from producing a termination email. The Court inferred that the evidence list had been generated using artificial intelligence and referred to documents that either did not exist, could not be identified or did not match the descriptions given to them.
Judge Symons described the proceeding as a cautionary tale about the dangers of relying on AI chatbots that present themselves as legal advisers. Her Honour emphasised that AI is not a lawyer, holds no practising certificate and owes no duty to the Court, the administration of justice or the person relying upon its output. The Court also observed that AI tools may prioritise producing an agreeable answer over providing an accurate one.
The applicant’s reliance on the misleading evidence list caused problems from the beginning of the proceeding. The Court found that the conduct arguably amounted to an abuse of process and came close to contempt of court, which can carry serious penalties. The claim was dismissed because of the applicant’s failure to comply with the Court’s orders, without the Court determining whether his underlying employment claim had any merit.
The result is particularly significant because the applicant may potentially have had an arguable claim. However, his reliance on AI-generated material and failure to verify the supposed evidence prevented that claim from being properly considered.
Together, Asif and Ba demonstrate that AI can do more than weaken a self-represented person’s presentation. It can cause the person to rely on arguments they do not understand, refer to evidence that does not exist, breach court orders, incur an adverse costs order or lose the opportunity to have a potentially valid claim decided on its merits.
Having AI produce a professional-looking document is not the same as receiving legal advice, understanding the evidence or being able to present a case in court.
You Must Be Able to Explain Your Own Case
A self-represented litigant cannot simply file an AI-generated document and expect it to speak for them.
A judge may ask why a particular order is sought, what evidence supports an allegation, how a legal authority applies or where the court obtains power to grant the requested relief. If the person does not understand the document or the legal principles it contains, they may be unable to answer those questions.
AI can also generate large amounts of irrelevant material. More words do not necessarily produce a stronger case. Repetitive or unfocused submissions may conceal the real issue, increase the other party’s costs and make it more difficult for the Court to understand the argument.
A lawyer’s role is not simply to produce formal-sounding documents. A lawyer identifies the legally relevant issues, evaluates the available evidence, advises about the strengths and weaknesses of the case, complies with procedural requirements and presents the client’s position in a way that assists the Court.
The Growth of AI-Generated Claims
The problem is becoming increasingly visible in Australian courts.
The Australian Financial Review recently reported that the number of self-represented litigants commencing workplace disputes in the Federal Court had more than doubled over the preceding year. The reported increase has raised concerns about a growing volume of poorly prepared or AI assisted claims.
AI may make it easier to generate and file a claim, but it cannot determine whether the claim has reasonable prospects of success. It may encourage a person to pursue an argument because the chatbot has presented it confidently, without properly identifying jurisdictional barriers, evidentiary problems or the risk of an adverse costs order.
AI Can Invent Cases and Misstate the Law
One of the most serious risks is an AI “hallucination”. This occurs when an AI tool creates information that appears plausible but is false.
In a legal matter, an AI hallucination may involve:
– a case that does not exist;
– a false quotation attributed to a real judge;
– an incorrect case name or citation;
– a legislative provision that has been repealed or never existed;
– a genuine case described as supporting a proposition it does not support; or
– law from another country presented as if it applies in Queensland.
In the decision involving the self-represented visa applicant, Judge Humphreys described testing ChatGPT by asking it about one of his own recent decisions. The program ultimately admitted that it had fabricated the judgment, the surrounding commentary and statements supposedly made by the judge.
The Queensland Courts similarly warn that AI-generated material can appear to come from a genuine source even when it is entirely fictitious.
The Court Will Hold You Responsible
A person remains responsible for every document filed under their name, regardless of whether it was written personally, copied from the internet or generated by AI.
Using ChatGPT is not an excuse for:
– making a false or misleading statement;
– relying on a non-existent authority;
– failing to provide admissible evidence;
– commencing proceedings in the wrong jurisdiction;
– missing a limitation period or court deadline;
– failing to comply with procedural orders; or
– seeking relief the court has no power to grant.
The Queensland Courts warn that where inaccurate or fictitious AI-generated material causes delay, the person responsible may face a costs order. In practical terms, attempting to save legal fees may result in the person losing their case and being required to pay some of the opposing party’s legal costs.
There may also be consequences beyond costs. A weak or incorrectly framed case could be dismissed before its merits are properly considered. Depending on the circumstances, the person may then be prevented from bringing the same claim again.
AI Cannot Give Independent Advice
AI usually responds to the way a question is framed. If a person presents only their own interpretation of events, the chatbot may reinforce that position rather than identify facts or legal principles that undermine it.
A lawyer is required to assess the matter independently. That may involve delivering advice the client does not want to hear, including that a proposed claim is weak, that particular evidence is insufficient or that settlement would be preferable to litigation.
That independent assessment can prevent a person from investing considerable time, money and emotion in proceedings that are unlikely to succeed.
Legal Advice May Cost Less Than Getting It Wrong
The attraction of AI is understandable. It is fast, inexpensive and capable of producing documents that appear professional.
The danger lies in confusing presentation with accuracy. A defective claim, missed deadline, disclosure of confidential information or adverse costs order can be considerably more expensive than obtaining legal advice at the beginning.
Even where a person does not engage a lawyer to conduct the entire proceeding, targeted advice may help them understand their rights, identify the correct process and avoid potentially irreversible mistakes.
Artificial intelligence can assist with legal work, but it cannot replace the professional judgment, accountability and tailored advice of an experienced lawyer.
If you are considering commencing proceedings, are already representing yourself or have used artificial intelligence to prepare a legal document, contact the experienced team at Miller Sockhill Lawyers on (07) 5444 4750.