ChatGPT and divorce in Quebec: why artificial intelligence does not replace a lawyer
More and more Quebecers are turning to ChatGPT for quick answers about their divorce. Calculating child support, dividing family assets, child custody — artificial intelligence seems to offer a tempting shortcut. But beware: using ChatGPT for your divorce carries serious legal risks that most people are unaware of. What you write to ChatGPT could even end up in front of a judge.
What ChatGPT Can and Can’t Do When It Comes to Divorce
ChatGPT is a generative artificial intelligence tool. It generates answers by predicting the most likely words following a question, based on millions of texts ingested during its training. He does not reason legally. He does not know your file. He does not check his sources.
In terms of Quebec family law, this poses several concrete problems. The Civil Code of Québec, the federal Divorce Act and the jurisprudence of the Superior Court of Québec are constantly evolving. ChatGPT does not have access to decisions rendered after its cut-off date and cannot apply the specific criteria that judges use to decide a family dispute.
For example, the calculation of child support is based on the Quebec model for determining child support, a specific form that takes into account the income of both parents, custody time and special expenses. ChatGPT cannot fill out this form correctly for you. Worse yet, they may give you a completely wrong amount — and you could base your trade on a wrong number.
Your conversations with ChatGPT are not confidential
Contrary to what many believe, your exchanges with ChatGPT do not benefit from any legal protection. Solicitor-client privilege does not apply to a conversation with a bot. This means that everything you write to ChatGPT — your income, your hidden assets, your frustrations with your spouse, your strategies — could potentially be leaked.
OpenAI keeps your conversations to improve its templates, unless you manually disable this option. In the context of a disputed divorce, the opposing party could theoretically request the communication of this data. Imagine your ex-spouse finds out that you asked ChatGPT how to hide income or minimize your assets — this information could be devastating in front of a judge.
AI “hallucinations”: made-up judgments
One of the most documented risks of ChatGPT is its tendency to invent legal references that don’t exist. This phenomenon is called the “hallucinations” of artificial intelligence. The tool generates names of judgments, file numbers, and even citations of judges that seem perfectly credible — but are entirely fictitious.
This is not a theoretical risk. Several courts in Canada and elsewhere have already sanctioned litigants and lawyers who filed documents containing legal references invented by ChatGPT.
Specter Aviation v. Laprade (2025) — Superior Court of Quebec
In this landmark decision rendered on October 1, 2025, the Superior Court of Quebec imposed a penalty of $5,000 on an unrepresented litigant who had used artificial intelligence to prepare his legal proceedings. The AI-generated content contained multiple fictitious legal references and fabricated case law. The court found that this conduct constituted a substantial breach of the rules of procedure. The Court did not condemn the use of AI per se, but insisted that caution and human verification are essential at every step.
Mata v. Avianca (2023) — New York Federal Court
This American affair has gone around the world. Two lawyers used ChatGPT to draft a court motion in a personal injury case. The AI invented six completely fictitious court decisions, with false citations and internal references. When opposing lawyers wanted to verify these judgments, they discovered that none existed. The judge fined the lawyers $5,000 and called their conduct bad faith.
Zhang v. Chen (2024) — Supreme Court of British Columbia
A lawyer in British Columbia has been reprimanded for including two “hallucinations” generated by ChatGPT in a court application. Although the judge found that the lawyer did not intend to mislead the court, he was ordered to pay the court costs personally. This decision is one of the first in Canada where precedents invented by ChatGPT have been identified in a court case.
Ko v. Li (2025) — Ontario Superior Court
In a case involving family and estate law, a legal brief contained fictitious case law presumably generated by artificial intelligence. The court made it clear that lawyers have a duty to ensure human verification of any material prepared by non-human technology, including generative artificial intelligence.
The Superior Court of Québec warns the legal community
On October 24, 2023, the Superior Court of Quebec issued a formal notice to the legal community regarding the use of artificial intelligence. The court reiterated that all legal claims must be supported by reliable and authoritative sources. This opinion is intended for both lawyers and litigants who represent themselves.
In September 2025, the Superior Court also published a governance framework on artificial intelligence, reinforcing its expectations for transparency and rigour in the use of these tools.
The Barreau du Québec has also issued clear warnings. Under section 128 of the Act respecting the Barreau du Québec, only a lawyer may provide personalized legal advice. ChatGPT can offer general legal information, but it can’t advise you on your particular situation — and the distinction is important.
Risks specific to divorce and family law
Family law is a field that is particularly ill-suited to artificial intelligence, for several reasons.
First, every family situation is unique. The division of the family patrimony, the determination of the compensatory allowance, the allocation of custody of the children — each of these elements depends on a complex set of factors that only a professional can properly assess. ChatGPT produces generic responses that don’t take into account your particular circumstances.
Second, artificial intelligence cannot detect situations of domestic violence, power imbalances between the parties, or issues related to child protection. A family law lawyer is trained to identify these situations and adapt their strategy accordingly. ChatGPT simply doesn’t have that capability.
Third, using ChatGPT to write messages to your ex-spouse can make the conflict worse rather than resolve it. AI-generated responses are often impersonal, sometimes awkward, and can be perceived as provocative in an emotionally charged context. In a child custody case, the tone of your communications is scrutinized by the court — messages that are visibly written by a machine could damage your credibility.
What to do if you’ve already used ChatGPT for your divorce?
If you’ve already shared sensitive information with ChatGPT as part of your divorce, here are the recommended steps. Disable chat history in OpenAI’s settings to prevent your data from being used for model training. Never submit a ChatGPT-generated document to a court without a lawyer first verifying it. Inform your lawyer that you have used AI — this transparency will allow them to assess potential risks to your case.
Above all, don’t make any important decisions based solely on ChatGPT’s answers. The calculation of child support, the division of property, custody arrangements — these will have a direct impact on your life and the lives of your children for years to come.
Artificial intelligence is a tool, not a lawyer
ChatGPT can be useful for getting general information about the divorce process in Quebec. They can help you understand some basic legal concepts, prepare questions for your consultation with a lawyer, or organize your ideas. But it is in no way a substitute for professional legal advice.
Quebec and Canadian courts are sending a clear message: the use of artificial intelligence in legal proceedings is permitted, but it requires rigorous human verification. The financial penalties imposed in the cases of Specter Aviation, Mata v. Avianca, Zhang v. Chen and Ko v. Li demonstrate that the consequences of careless use are very real.
It is also important to note that family law laws vary greatly from province to province in Canada. Quebec has its own Civil Code, distinct from the common law that applies in the rest of the country. ChatGPT does not always make this distinction and may provide you with information based on Ontario or U.S. law without notice to you. Erroneous legal advice due to the wrong jurisdiction could have catastrophic consequences on the outcome of your case.
In addition, artificial intelligence algorithms do not understand the nuances of negotiation in family law. A good lawyer knows when to make strategic concessions, when to emphasize certain points, and how to structure an agreement that protects your long-term interests. ChatGPT gives you a static answer, disconnected from the real dynamics of your case and the practices of the Quebec courts.
Before entrusting your divorce to a robot, consult a lawyer. Your situation deserves a human, personalized, and privileged analysis — something ChatGPT will never be able to offer.
Consult a family law lawyer
Are you going through a divorce or separation in Quebec? Don’t let artificial intelligence make decisions that will affect your future and that of your children. At Azran Lawyers, our family law lawyers understand the realities of divorce in Quebec and can accompany you every step of the way. Whether you have questions about child support, division of family assets, child custody or any other matter related to your separation, we are here to help.
Contact us today for a consultation. Professional legal advice can save you from costly mistakes — and unpleasant surprises in court.
