GenAI is changing how self-represented parties engage with Canada’s courts

In the fall of 2018, a man arrived in Canada from Nigeria and filed a claim for refugee protection. The man later told immigration authorities he risked facing persecution by the Nigerian government and Black Axe, a criminal organization, if he returned home. After his application was rejected, he filed an appeal in court.

In British Columbia, another litigant was embroiled in her own legal battle. In the spring of 2025, she had been arrested by police officers and detained at a local hospital. The woman, who is transgender, sued the province and a provincial health authority, arguing she was unlawfully detained and treated in a degrading and discriminatory manner. 

Around the same time, separated couples were fighting in courts in Alberta and Saskatchewan over how to divide time with their children. Another woman in BC asked a court to seal records related to her workers’ compensation claim. A man in Quebec, sentenced to community service and years of probation after being convicted of assault and uttering threats, appealed his conviction and sentence.

These cases have two things in common. First, each of the litigants – with the exception of one of the parties in the Saskatchewan family law case – represented themselves in court, without the help of a lawyer. Second, like a growing number of people who cannot afford to hire a lawyer but find themselves mired in Canada’s complex legal system, they turned to a modern solution to help them make their case in court: artificial intelligence.

Self-representation can be daunting. In court, self-represented litigants are expected to follow the same rules, deadlines, and procedures as lawyers, without the benefit of the formal training or years of experience that lawyers possess. But self-representation is not a new phenomenon, or even a rare one. In 2016, the federal government noted that the number of self-represented litigants across Canada was increasing, citing studies showing it had been climbing for the previous two decades.

In March, Statistics Canada said that in 65 percent of family law cases that were active in 2024 and 2025, at least one party reported being self-represented at some point during the case. One of the Supreme Court of Canada’s most high-profile rulings this year, in Ahluwalia v. Ahluwalia, involved a litigant who represented herself for much of her legal battle against her husband, whom she accused of domestic abuse. The high court has reported that 37 percent of the leave applications it received in 2025 were filed by self-represented litigants.

In recent years, the growing availability of generative AI tools has spawned a new set of statistics: according to websites tracking court and tribunal decisions that address fake citations or arguments created by AI, the vast majority of the cases involve self-represented litigants, rather than lawyers, submitting hallucinated materials. Many people across Canada are clearly using AI to help draft materials for their cases. At the moment, there is not enough data to paint an accurate picture of how widespread the phenomenon is.

But courts have noticed a shift on an anecdotal level. In response to a Canadian Lawyer survey about how Canada’s courts are engaging with generative AI, several courts flagged an apparent increase in the number of self-represented litigants using AI to prepare court submissions. Filings look different than they used to, litigants are filing more materials, and they’re more convinced that they’re right. The growing availability of AI tools has changed how self-represented litigants engage with the courts, and courts are in the midst of figuring out how to adjust.

To learn more about this shifting landscape, Canadian Lawyer talked to Superior Court of Quebec Chief Justice Marie-Anne Paquette, Justice Steven Hinkley of the Alberta Court of Justice, British Columbia Court of Appeal legal counsel Shirley Smiley, and Jennifer Leitch of the National Self-Represented Litigants Project. Below, they detail how AI has impacted their experiences with self-represented litigants, and what they believe needs to change to improve those litigants’ access to justice.

Improved formatting, same substance

In separate conversations in July, Paquette and Hinkley each responded to a question about what trends they’ve observed on self-represented litigants’ AI use by emphasizing the same point: many people have successfully used AI to raise the quality of their court submissions.

“This is not a trivial benefit, either for the litigants or for the judges,” Paquette says. She notes that since generative AI tools became widely available, many filings by self-represented litigants are “clearer, better organized, and closer to the expected structure of legal proceedings than what we would have seen a few years ago.” Hinkley puts it another way.

“The format gap has closed,” he says. Different legal documents need to be structured in distinct ways, depending on their purpose, the court they’re being filed in, and procedural rules. Hinkley has seen many breaches of formatting rules during his time on the bench, from materials that were handwritten to submissions that were clearly copied and pasted from the internet.

“Materials from self-represented litigants now look like what they should look like,” he observed. “There’s some organization to the affidavits. The applications we’re seeing from self-reps, or at least I’m seeing from self-reps, are recognizable as what they are supposed to be, and that’s a real change.”

However, Hinkley says the improved formatting of many filings has not changed the quality of the actual arguments they’re making. “The self-represented party I’m seeing has the same command or lack thereof of their cases they did before,” he says. Hinkley notes that the wider availability of generative AI tools has also correlated with a greater difficulty in convincing self-represented litigants to let go of bad arguments; this may be due, he speculates, to the sycophantic nature of many AI chatbots, which are likely telling litigants that their arguments are sound.

Paquette’s experience is similar. While formatting has improved in recent years, she and other judges at the Superior Court of Quebec have seen an increase in fake case citations or references to irrelevant authorities. Many of these filings appear polished or have attractive legal reasoning, but lack actual substance, she says.

She’s also noted another phenomenon. Because AI tools dramatically lower the cost and effort involved in producing documents, the length of filings by many self-represented litigants has increased. “This places an additional pressure on judicial resources, a trend that we can already observe very, very precisely and concretely in my office,” Paquette says.

She offered an example. At the Superior Court of Quebec, the chief justice is responsible for deciding whether a person who has been deemed a vexatious litigant can launch court proceedings. Last year, applications by vexatious litigants for permission to initiate proceedings “have become noticeably longer and more sophisticated,” Paquette says. For these litigants, who typically earn their distinction by repeatedly filing meritless lawsuits to harass other parties, generative AI tools can be damaging, she argues, adding these tools multiply “by 10 the waste of time and pressure they can put on the system.”

Smiley said the BC Court of Appeal is facing similar concerns. “Because genAI platforms make it easier for a litigant to produce more material, faster, and at little or no expense, litigants may file more applications or apply to review or vary previous decisions even with very little chance of success,” Smiley wrote in an email.

“This burdens the court and responding litigants without improving access to justice outcomes.”

The verification burden

For Paquette, one of the biggest challenges resulting from these trends is how they’ve exacerbated the so-called verification burden – the obligation of the courts to ensure that each case or authority cited by parties is real and accurate, rather than, say, an AI-hallucinated invention.

“When a judge or a law clerk cannot take the quotes or the references at face value, every authority in a filing must potentially be checked,” Paquette says.

“That work… existed before, but the risk profile has changed,” she adds. “A fabricated quote is more insidious than a weak argument because it looks legitimate. And the court staff also faces documents that are more voluminous and superficially well-formed, which makes the initial triage more demanding rather than less demanding.”

Before generative AI tools were widely available, Hinkley says, “I’d read a brief, I’d follow the argument, and as long as the argument made sense, I could follow the logic, and I could see it cited materials that I was familiar with, that was the job.

“Now there’s a second question that runs… do the cited authorities actually say it? Do they exist? Do they mean what the person who’s advancing them thinks they mean?” Hinkley says. This is less of a concern with lawyers than self-represented litigants, who may not understand that they’ve produced inaccurate or nonexistent citations, the justice says.

“What it boils down to really is this: reading submitted materials, particularly from self-represented litigants, used to be one job,” Hinkley says. “Now it’s two.”

For Paquette, a single instance of failing to meet the verification burden can have far-reaching consequences. While lawyers and self-represented litigants are ultimately responsible for the accuracy of their own submissions, a court ruling that relies on fabricated material would have a dramatic impact on the public’s trust in the justice system, the chief justice says.

“For this reason, the judges and the courts are extremely cautious and careful,” Paquette adds.

Providing education

While the Superior Court of Quebec’s judges and staff have noticed a palpable shift in how self-represented parties engage with the court since generative AI tools became broadly available, Paquette argues these observations are “necessarily impressionistic” because there is currently no way to confirm when a filing has been prepared with AI.

She notes that several courts require parties to disclose when they’ve prepared materials using AI. However, many litigants have ignored these requirements or failed to accurately report their AI use because they did not understand that they had used an AI tool.

As executive director of the National Self-Represented Litigants Project, Leitch says her organization, which provides support to people representing themselves in court, has faced a similar challenge. On the NSRLP’s website, self-represented litigants seeking resources typically fill out an intake survey. In April 2024, the organization added questions about AI use to the survey. As of June 2026, 51.4 percent of respondents had reported using AI tools to help represent themselves in court.

“What we saw was a… very slow but steady increase in the percentage of self-reps that were acknowledging that they’re using AI,” Leitch says. “I would have thought there would have been a much more significant explosion of AI use, like Claude or ChatGPT, by self-reps.” Leitch notes, however, that responses to qualitative questions in the survey indicate that respondents may be underreporting their AI use based on a misunderstanding of what exactly constitutes AI.

To assist those individuals who plan to use AI to advance their case, the NSRLP have hosted webinars explaining how generative AI tools work, including their potential for producing hallucinations or high volumes of irrelevant material. The webinars also gave directions for prompts and explained basic legal concepts, like how different jurisdictions have different laws.

Paquette agrees that a key component of helping self-represented litigants navigate AI tools is education.

“Litigants need to understand that a general purpose chatbot is not a source of law, and that it can invent authorities with complete confidence, and that they remain responsible for what they file,” she says, adding that plain language guidance from the courts can go a long way.

While Hinkley agrees, noting an encouraging trend of Canadian courts publishing plain language summaries of court decisions or court processes, he argues that at the end of the day, what self-represented litigants need most is to know “whether or not what they’re advancing and what they believe is correct for the matter that they are dealing with.

“That has been a challenge for self-represented litigants for as long as there have been self-represented litigants,” he says.