Can I just use AI to do my divorce?
You can ask AI questions. You shouldn't rely on it for your filings. Public AI tools — ChatGPT, Gemini, Claude, Copilot, and the rest — were built to predict the next plausible word, not to know New Hampshire law. They confidently invent court cases, get filing deadlines wrong, and don't know that your local family division circuit court has its own rules. Researchers have logged more than 1,200 fake legal citations that AI tools have invented and that people then filed in court — and about 59% of those came from self-represented filers, not lawyers.
The pattern is consistent: AI gives a confident-sounding answer, the reader trusts it, the court catches the made-up case, and the filing gets dismissed or the filer gets sanctioned. We've seen a self-represented plaintiff hit with $2,632 in attorney's fees and a case dismissed with prejudice for two fabricated citations in a brief. We've seen lawyers fined $5,000 (Mata v. Avianca, Inc., S.D.N.Y. 2023), $10,000 (Mostafavi, California, 2025), and $110,000 (Oregon federal court, 2026 — the largest such penalty on record) for the same mistake.
Divorce Tutor was built by a NH family law attorney with 25+ years of practice. The questions, forms, and guidance are written by her — not generated by AI — and the documents you produce are court-ready New Hampshire forms. That's the difference.
Are my conversations with AI private? Can they be used against me in my divorce?
No, they're not private. Public AI tools — ChatGPT, Gemini, Claude, Copilot, and the rest — are not your attorney and not your therapist. There's no attorney-client privilege, nothing protecting what you type into them. In 2026, a New York court formally confirmed that AI chat logs are discoverable — meaning the other side's lawyer can subpoena them and use them in your case.
That matters because of what people type. If you ask an AI something like "can my husband make a claim against my $500,000 inheritance from my mother?" — you've now created a written record disclosing the inheritance, the amount, and that you were trying to figure out whether to hide it. Opposing counsel would love to put that in front of a judge.
If you want to talk through your case, talk to a real attorney (covered by privilege) or use a platform like Divorce Tutor, where your answers stay in your private account instead of being fed into a public AI that learns from what you type.
Did you use AI to write the questions and content on Divorce Tutor?
No. Every form question, tip, video script, and explanation was written by Heather, the NH family law attorney who built the platform — drawing on 25+ years of practicing in New Hampshire family court. The forms are real NH court forms. The workflows reflect actual templates used in the same courts in which you will file. The guidance reflects what Heather would tell a client in her office.
We use software to deliver that content — that's how you can fill out a Petition for Divorce online instead of by hand — but the words are hers.
How is Divorce Tutor different from a general AI tool?
Three things:
- The content was written by a NH family law attorney, tied to actual NH court forms, rules and NH law. A general AI doesn't know your local court's rules, won't catch that the filing fee changed in July 2025 or that the Parenting Plan changed in January 2025, and will confidently make up a deadline.
- The form generated is a NH court-ready document. You answer plain-language questions; Divorce Tutor produces the actual form based on the NH Judicial Branch forms that the court accepts. An AI chatbot produces text that is not properly formatted for filing with the court.
- Your information stays private. Your answers live in your account, behind your password. Unlike a public chatbot, they aren't fed into an AI that learns from them — and they aren't discoverable just because you typed them.
If you want to brainstorm or learn vocabulary, AI is fine for that. If you want to file something with a New Hampshire court, that's what Divorce Tutor is for.
What actually happened to people who used AI to file court documents?
These are real court rulings, not cautionary tales. In each one, someone trusted what an AI tool produced, filed it, and the court caught it:
- New York (2026): A court sanctioned an attorney and a law firm for filing a brief — prepared with generative AI — that cited cases that don't exist and quoted fictitious "Court of Appeals" rulings that were flatly contrary to actual law. (Landberg v. City of New York, 2026 NY Slip Op 03935, Docket No. 2025-02380, June 23, 2026 — Justia U.S. Law)
- New York federal court (2024): A judge refused to credit evidence a firm submitted using ChatGPT to show its billing rates were reasonable, because "ChatGPT has been shown to be an unreliable source." (Z.H. v. N.Y.C. Dep't of Educ., No. 23-cv-3081 (ER), 2024 U.S. Dist. LEXIS 124478 (S.D.N.Y. July 12, 2024) — American Bar Association)
- Federal appeals court (2024): The court referred an attorney to the Second Circuit's Grievance Panel for investigation after he filed a brief that relied on "non-existent" caselaw generated by ChatGPT. (Park v. Kim, 91 F.4th 610 (2d Cir. Jan. 30, 2024) — American Bar Association)
- Wisconsin federal court (2024): The court found that a case cited in a reply brief appeared to be hallucinated by AI — it couldn't be found anywhere — and warned counsel they would be sanctioned for any future non-existent citations. (Plumbers v. Morris Plumbing, LLC, No. 23-CV-616-JPS-JPS, 2024 U.S. Dist. LEXIS 70751 (E.D. Wis. Apr. 18, 2024) — American Bar Association)
- Mississippi (2026): Lawyers used AI-generated material containing false information, leading a federal judge to impose sanctions and fines. ("Judge Boots Attorneys After Both Sides Submit AI-Hallucinated Citations," Forbes, June 12, 2026)
- Wyoming (2026): A federal judge threatened to sanction two lawyers at Morgan & Morgan who included fictitious citations in a lawsuit against Walmart; one admitted he had used an AI program that "hallucinated" the cases and apologized for the mistake. (Law.com, May 27, 2026; see also Reuters, May 29, 2026)
- Federal appeals court (2026): The Seventh Circuit imposed a $5,000 sanction on a litigant's counsel for briefs containing numerous fabricated citations and factual misrepresentations produced by AI tools, and referred a second attorney to the Illinois disciplinary authorities. (Perez-Castillo v. Blanche, No. 25-1988 (7th Cir. 2026) — Justia U.S. Law)
The pattern never changes: the AI sounds confident, invents a case or a quote, and the court catches it. And it falls hardest on people without a lawyer — researchers have documented more than 1,200 fabricated AI citations filed in court, and most came from self-represented filers. As of 2026, hundreds of federal judges have standing orders requiring filers to disclose AI use, and family courts are catching up fast.
