People Are Using AI for Divorce Advice and It’s Driving Their Lawyers Crazy | Hindustan Times

In a spring survey of 133 lawyers at Cordell & Cordell, a national family-law firm, 80% said clients had sent them correspondence, draft pleadings or strategy suggestions they believed were generated with AI. Feed one side of a family fight into a chatbot and you might get the legal answer you want to see, but it won’t necessarily fly with the judge.Steven Benmor, a Toronto family lawyer, recently drafted a prenuptial agreement the way he said his client asked. Then the client emailed him a 13-part list of revisions nearly as long as the prenup.

Some revisions went so deep into the weeds, Benmor said, that no human could have come up with them. Benmor knew his client had turned to an artificial intelligence chatbot. And he knew that taking the time to explain why the chatbot was mistaken would require charging his client extra.

His mantra: “Don’t spend money unduly by asking a human expert to enter into a debate with a robot expert.”

It was hardly the first time this had happened to Benmor. Family lawyers, who handle everything from prenups to divorces and custody fights, say many of their clients are now showing up with pages of legal advice that look like the result of ChatGPT or other chatbots.

In a spring survey of 133 lawyers at Cordell & Cordell, a national family-law firm, 80% said clients had sent them correspondence, draft pleadings or strategy suggestions they believed were generated with AI.

People ask chatbots for everything these days, including financial advice and medical diagnoses. But family law isn’t a place where they shine, lawyers say. Feed one side of a family fight into a chatbot and you might get the legal answer you want to see, even if it isn’t one that will necessarily fly with the judge.

Divorcing partners turn to them anyway. Money is often tight for people when a marriage comes apart, and hiring a lawyer takes time. A chatbot answers right away, even in the middle of the night.

Logan Burns used ChatGPT for legal advice and ended up having to apologize to the judge. Burns, a 35-year-old pastor and roofing salesman in LaGrange, Ga., paid a Georgia lawyer a $2,500 retainer when his marriage came apart last year. The couple called off the divorce and tried to reconcile, and his wife took the children to stay with family in Montana. When the reconciliation failed, the custody fight landed there, where a lawyer wanted $7,000. He paid $19.99 a month for the chatbot instead.

“It was doing the job of a lawyer, or at least a paralegal,” he said.

ChatGPT pointed him to a rule that generally lets a state decide custody once a child has lived there for six months. His children had been in Montana for less than that. Burns wanted the case moved to Georgia.

So he wrote the judge a letter arguing that the court had improperly taken over his children’s case. The judge brought it up in court. “You technically don’t have jurisdiction,” Burns recalled telling the judge. The judge sternly told him that wasn’t how it worked. The rule has exceptions: Courts can step in during emergencies, and Burns had filed an emergency request himself.

“Your Honor, I’m not accusing you of anything,” he recalled saying. “ChatGPT gave me the baseline information. But it doesn’t give you the interpretation of law.”

He has temporary custody of his children, though the case is ongoing. “Thank you, AI,” Burns said.

Damien Charlotin, an adjunct professor at UC Law San Francisco who tracks AI-generated errors in court cases, has counted 148 family-law rulings in which judges flagged fake citations or other apparent AI mistakes. More than 100 came this year.

Chatbots also tend to take one side’s word at face value, even though two sides often disagree on facts in family law. The owner of an online wellness company had agreed in his divorce to pay hefty spousal support, based on what the business was earning. Years later, taking far less money out of the business, he asked a chatbot whether a judge would cut his payments. The bot told him he had an easy win.

The chatbot never asked to see his financial statements, or his spending, which hadn’t shrunk with his income. It “simply accepted our client’s narrative regarding his finances,” said Hossein Berenji, a Los Angeles family lawyer whose firm later took the case.

Relying on the chatbot’s advice, he hired a lawyer and persuaded the attorney to follow the chatbot’s strategy. In late 2024, the lawyer asked the court to reduce the support, attaching some company records. The other side’s lawyer used those same records to argue his income hadn’t really fallen, and about a year later the judge agreed. The man blamed his lawyer, not the chatbot, and switched to Berenji’s firm. “It took us a while to convince him to move away from what AI was advising him to do,” Berenji said.

The firm brought in a forensic accountant, who found what the first filing had left out: A lot of the company’s products had spoiled while the business was still paying its overhead, Berenji said. Sales were big. What was left for his client wasn’t.

At a recent hearing, a judge tentatively agreed his finances had changed and put his support payments on hold for now.

What the chatbot called an easy win took two law firms and nearly two years.

Everything the man paid his first lawyer was “basically money thrown down the drain,” and he likely overpaid spousal support the whole time, Berenji said.

Maha Kamal, a Denver family lawyer, joined a custody case on behalf of a mother after the children’s father began peppering the court with filings that accused her of breaking the judge’s orders. The judge would reject one, and another would arrive. The father was mostly representing himself.

The filings appeared to be written by AI, Kamal said. “It was scaring my client.”

Instead of answering them one by one, Kamal asked the judge to call both parents in. By then, there were nine filings. She pointed out that one cited a legal principle that didn’t exist. The judge warned the father that if it happened again, he could be ordered to pay the mother’s legal fees, among other penalties, Kamal said.

He denied using AI. He also argued that nothing prohibited it, she said.

He kept filing anyway. The judge kept turning him down and said the court wouldn’t consider more filings before the final hearing. The parents eventually settled.

“It could be used as a form of litigation abuse,” Kamal said.

It isn’t just the lawyers losing patience. Judges are as well.

A Florida husband representing himself in his divorce, who used AI, filed at least 145 motions, then sent an appeals court 45 AI-generated pages and an 863-page appendix explaining himself, according to court records.

In May, the court ordered its clerk to reject his future filings unless a lawyer signs them.

Write to Dalvin Brown at dalvin.brown@wsj.com