TL;DR
A new report in England and Wales warns that parents are turning to AI for legal information while effective safeguards and routes to challenge errors remain limited.
Research shows people are already using chatbots for legal problems, often alongside other sources of help.
The question isn’t whether a chatbot can make legal paperwork easier. It’s who checks the answer before it affects a family.

I can understand why someone facing a family court dispute would ask a chatbot for help.
Legal language is difficult. Lawyers cost money. You might be trying to understand a form, a deadline or a letter you’ve just received, while already dealing with a relationship that has become painful and complicated. You type a question and get an answer in seconds, in ordinary English.
That can be helpful..
But the answer might be wrong. And when it comes to a dispute involving children, the mistake can travel further than the conversation in which it was made.
On November 2, the Nuffield Family Justice Observatory published a report by Dr Natalie Byrom about privately owned technology in the family justice system in England and Wales. Its concern isn’t simply that AI sometimes makes things up. It’s that people and courts are increasingly relying on tools without equally clear ways to evaluate them, oversee their use or challenge harmful errors.
I’ve written before about what happens when AI keeps telling someone they’re right. This report raises another question. What happens after the chatbot’s answer becomes part of what someone believes, writes down or presents to a court?
People are already asking
A survey by the UK legal charity JUSTICE and the Administrative Fairness Lab questioned 3,287 people between December 2025 and January 2026. Of the 1,428 respondents who said they’d experienced a legal dispute in the previous two years, 233 said they’d used an AI chatbot to discuss it.
That works out to about one in six people in that subgroup (not one in six separating parents.) The disputes included housing, employment, debt, consumer matters and family law. The survey doesn’t establish how many family-court litigants use AI.
Just as important, only 6% of the people who used AI said they’d relied on it alone. Many looked elsewhere too: search engines, government websites, advice organizations or people they knew.
The researchers also examined 77 shared chatbot transcripts. People weren’t all asking for a legal strategy. Some wanted confusing language explained or help writing a message. Others wanted to know what to expect in court. Those are different needs, and some are more suited to a language tool than others. The study describes how people used the tools, not whether the answers were legally accurate.
There is a reason the technology is attractive. When someone can’t afford professional help, a fast explanation may be the first understandable answer they’ve received.
That benefit matters. So does the difference between an explanation that sounds right and one supported by the actual law and facts.
When the answer becomes a document
The Nuffield report describes a system in which private technology providers increasingly supply tools used by the public and by justice institutions. Some AI-assisted information and document-preparation tools operate outside the regulation that applies to professional legal services.
There are two different situations here. A parent using a public chatbot to understand a form or prepare a statement is responsible for deciding what to submit, often without affordable legal help. If the advice gets a deadline or legal requirement wrong, that parent may not discover the problem until after acting on it. The report raises questions about whether consumer-facing tools have adequate checks and whether people have a practical way to challenge mistakes.
A court or public agency using AI to summarize documents is a different matter. The institution has chosen the tool and the way it’s used. People affected by its output may need to know whether the original documents were checked, who reviewed the summary and how they can correct an error. The report doesn’t establish that a particular family court decided a case incorrectly because of AI. It does make the case for independent evaluation and clearer accountability before such tools become routine.
In When the Messages Became Evidence, I wrote about an Indiana case where photographs of text messages mattered in a dispute over college expenses. The court didn’t establish that every message was delivered or read. It considered the admitted material, the testimony and the specific legal issue before it.
That piece concerned actual records and how they were challenged. The new problem is different: an AI draft may arrive with the appearance of a finished explanation before anyone has tested the underlying claims.
And there is an important danger in getting this wrong. Someone describing abuse may have very good reasons for being afraid or refusing contact. Someone who believes a child has been influenced may also have genuine concerns. A chatbot can’t settle those competing possibilities simply by turning one account into better prose.
Who gets to challenge the mistake?
Byrom’s report makes 26 recommendations in four broad areas. They include clearer regulation of consumer AI legal tools, better public legal information, independent evaluation of AI used in courts and safeguards that let people contest its use.
The report says judges in England and Wales have been given access to a secure version of Microsoft Copilot, while public information about its approved tasks, evaluation and monitoring remains limited. That is a transparency concern raised by the report, not evidence that judges have been using it improperly or making AI-driven decisions.
For a parent using a consumer chatbot, accountability might mean a clearer warning about what the tool cannot do, a reliable source for the applicable law and somewhere to report an error. None of those measures substitutes for legal advice. They do address the problem of a person being left with a confident answer and no practical way to check it.
When a court uses AI, the questions change. Was the tool tested for the task? Can someone compare its output with the original material? Will the people affected know AI was used and have a way to challenge a consequential mistake? Simply saying a human reviewed the result doesn’t answer those questions. The report’s recommendations are aimed at that larger problem of oversight, not just at improving chatbot wording.
This isn’t a claim that every use of AI in family justice is unsafe. The Nuffield report itself argues for realizing the technology’s benefits under better standards. It also focuses on England and Wales. Its findings cannot be treated as a description of the rules, court practices or available remedies in every American state.
What courts should make clear
I’ve spent a lot of time writing about records and the limits of what records can establish. They can help show what was said, when an attempt at contact happened or what someone believed at the time. They don’t automatically explain another person’s private experience.
If a court or public agency uses AI to handle material that affects a family, I’d want the original documents available for comparison and a clear answer about who checked the result. I’d also want the people affected to know how to challenge a mistake. That’s a different responsibility from a parent deciding whether to trust advice from a chatbot at home.
I don’t think the answer is to tell families who can’t afford a lawyer to stop using tools that help them understand the process. The problem is leaving them to work out, on their own, when the tool has crossed from explaining the process into supplying an answer it cannot support.

