A person with no lawyer can sit down tonight with ChatGPT and produce a competent-looking first draft of a custody modification petition. Michelle May O'Neil thinks that is real progress. She also thinks it is about to create a problem the courts are not ready for.
"If you can afford a lawyer, the lawyer recognizes the error and catches it. If you can't afford a lawyer then nobody does," she wrote.
O'Neil is an appellate family lawyer in Frisco and the founder of Concierra Legal. The Best Lawyers in America lists her in its 2027 edition in both appellate practice and family law, and dates her recognition there to 2016.
She answered our questions in writing, exclusively for frisco.city. The questions below are ours, and the answers are hers, in her own words, edited for length and clarity.
The gap that replaces the gap
You started an Institute for Constitutional AI Law. What's the concrete problem you set it up to work on?
Access to justice is the concrete problem. AI is about to make it both better and worse at the same time.
Right now, roughly 80 percent of family law litigants in this country don't have a lawyer because they can't afford one. This means in large part they are lacking access to legal information and AI promises to close that gap and in many ways it will. A person can sit down with ChatGPT tonight and get enough information to create a decent first draft of a custody modification petition. Five years ago that person had nothing. So, that's real progress.
Here's what goes wrong, though. That petition may cite a statute that was amended last session. It may use language that waives a right the person doesn't know they have. It may sound perfectly professional and yet be substantively wrong.
Now scale that. If you can afford a lawyer, the lawyer recognizes the error and catches it. If you can't afford a lawyer then nobody does. The technology that was supposed to level the playing field just created a new gap between people who can afford to verify what the AI told them and people who can't.
The Institute exists to work on the constitutional dimension of that problem. What does due process look like when AI is generating the evidence or drafting the pleadings? What does equal protection mean when the quality of AI-assisted legal help tracks with income? These aren't hypothetical questions. They're in courtrooms right now, and nobody has built the framework to answer them.
Seven layers of identity
What is the Identity Stack, in plain terms, for someone who isn't a lawyer?
You have seven layers of identity, and most people only think about two or three of them.
The first is biological: your face, your voice, your DNA. The second is legal: your name, your Social Security number, your citizenship. The third is economic: your credit history, your employment record, your financial footprint.
Most people stop there. But there are four more.
Your representational identity is your image: photographs, video, how you appear to the world. Your algorithmic identity is the profile that data systems have built about you from your behavior: what you click, what you buy, where you go. Your synthetic identity is what AI can generate that looks and sounds like you: a deepfake of your face, a clone of your voice. Your agentic identity is an AI system that acts on your behalf: makes purchases, sends messages, negotiates.
Each of those layers can be created, copied, sold, or manipulated, and right now the law protects some of them and ignores others. The Identity Stack is a framework for seeing all seven at once, so that when you're drafting a contract, or a prenuptial agreement, or a piece of legislation, you can ask: which layers does this cover, and which ones did we miss?
I debuted it at SXSW 2026 in a talk called "Who Owns Me: The Legal War Over Identity in the AI-Driven Creator Economy," and it's now taught in curriculum at the Stanford Graduate School of Business.
The case she says changed Texas law
On Sept. 23, 2022, the Supreme Court of Texas conditionally granted a petition for writ of mandamus in In re Mariam Ayad, holding in a per curiam opinion that Sections 6.6015 and 153.00715 of the Family Code required a trial court to hear and determine challenges to a premarital agreement's validity and enforceability before referring the dispute to arbitration. The Court did not decide whether the agreement itself was valid. That, it said, is for the trial court to try in the first instance. O'Neil was among the lawyers for the relator, as legal scholar Eugene Volokh noted at the time.
The holding turns on Texas procedure rather than on which tribunal was named, and applies to any arbitration clause inside a challenged premarital agreement.
Pick whichever mattered most: In re Ayad, Moore v. Moore, Kausland v. Volek. What did Texas courts do differently afterwards?
In re Ayad. Not close.
A woman signed a premarital agreement that included a clause requiring arbitration of any divorce dispute before a Fiqh panel — an Islamic religious tribunal. When she filed for divorce, the husband moved to enforce that clause. The trial court ordered her into religious arbitration while ignoring all of the reasons why the agreement should not be enforced.
We took it to the Texas Supreme Court on mandamus. The briefing raised due process, the Establishment Clause, public policy, and women's equality. The Court granted the petition and directed the trial court to withdraw its order and required the trial court to hear her challenges to the agreement's validity first.
Before Ayad, a trial court could look at an arbitration clause in a premarital agreement and enforce it without first deciding whether the agreement itself was valid. It matters because the defenses to a premarital agreement — duress, unconscionability, involuntary execution — go to whether the person had a meaningful choice. If you skip that question and send someone straight to arbitration, you've assumed the answer. The Ayad case says you can't do that.
Moore v. Moore is the other one worth knowing. That's the leading modern Texas authority on setting aside a prenuptial agreement for involuntary execution and duress — a wife who signed under conditions that didn't amount to a free choice. And Kausland v. Volek, decided this year, answered a question that comes up in every prenup with a forfeiture clause: do you forfeit your benefits under a prenup if you challenge the agreement's validity directly? Or can you forfeit simply by asking for something the agreement prohibits? The court held it's the broader reading — requesting relief inconsistent with the agreement's terms triggered forfeiture. I represented the husband and prevailed.
What the Supreme Court sees that a trial court does not
Fifty-five cases before the Supreme Court of Texas is a lot of time watching that court. What does family law look like from up there that it doesn't from a trial bench?
One clarification first, because precision matters: fifty-five matters before the Supreme Court of Texas, not fifty-five argued cases. Most Supreme Court work is briefing in the form of petitions for review, merits briefs, mandamus petitions. I've presented three oral arguments before the Court. The distinction matters because the work is persuading a court that already knows the law to look at a problem differently, all via written format.
