AI in the Practice of Family Law

Artificial Intelligence in the Practice of Family Law: A Useful Tool with Ethical Traps

BY CHRISTOPHER R. CASTELLANO, ESQ.

WHEN YOU GOOGLE (OR, FOR THAT MATTER, USE YOUR PREFERRED ARTIFICIAL INTELLIGENCE (AI) platform) “AI content,” you will find scores of outlets reporting on a stark finding that has developed in the last year, namely, society crossing the tipping point where more online content is generated by artificial intelligence than humans. This new reality should rightfully give you pause and make you consider its wider implications. Of course, cautionary tales and articles about the use of artificial intelligence may be as plentiful online as AI-generated content, but a polarized approach to AI is simply not logical. Artificial intelligence is here, and it is not going anywhere.

In the legal world, we need to appreciate the benefits of AI whilst recognizing the risks and learn where each of us can strike that balance because, whether we like it or not, our clients use it to draft contracts, organize financial documents, summarize communication, and generate case summaries before they speak with counsel. Similarly, as one attorney uses AI to help prepare deposition outlines, summarize discovery, compare agreements, and refine correspondence, their opposing party/counsel may use it to create polished but inaccurate narratives. With all litigants having access to tools capable of fabricating or altering texts, emails, photographs, audio recordings, financial summaries, and other evidence, the entire legal community must rise to the challenge of understanding AI, including its strengths and weaknesses.

As a family law attorney in Maryland, I have found that the question of AI has become an ever-present factor, from pro-se parties providing AI-drafted custody agreements to the nowfamiliar AI-formatted summary of a document. Therefore, the question is how Maryland family law practitioners can use AI responsibly while recognizing the evidentiary and ethical risks it poses. Used carefully, AI can be valuable for organizing, drafting, issue spotting, and preparation. Used carelessly, it can mislead the court, compromise client confidences, distort negotiations, and damage the lawyer’s credibility.

AI-Generated Material and the Evidence Problem

Family law litigation is especially susceptible to AI-related evidentiary problems because the evidence often consists of ordinary communications and personal records, including text messages, screenshots, social media posts, photographs, spreadsheets, and chronologies of alleged facts, just to name a few. The problem is not limited to dramatic deepfakes. The more common risk is subtler.

A party uses AI to summarize communications, but the summary omits context. A party creates a spreadsheet using AI-assisted categorization, but the categories are wrong. A litigant “reconstructs” text messages from memory and produces them in a screenshot-like format. A parent submits an AI-polished chronology that blends facts, assumptions, and advocacy. The danger is not merely that AI can create false evidence. The danger is that AI can make unreliable material look organized, neutral, and authoritative.

Md. Rule 5-901 requires authentication or identification as a condition precedent to admissibility. In Sublet v. State, 442 Md. 632 (2015), Maryland’s high court addressed authentication of social media evidence and emphasized that the trial judge must determine whether there is proof from which a reasonable juror could find that the evidence is what the proponent claims it to be. That framework is certainly relevant to AI-era family law practice.

Just as a screenshot is not self-proving because it looks like a text message, a social media post is not automatically authentic because it bears a party’s name. When analyzing potentially probative information for its evidentiary value, family lawyers should ask practical foundation questions to determine the document’s creator, source, and whether it was edited, enhanced, summarized, translated, or regenerated. One valuable piece of information that is all too familiar but remains no less important in the ‘AI-era’ is metadata. An understanding of metadata can help to gain insight into these questions and help verify the authenticity/ credibility of information.

Artificial Intelligence in the Practice of Family Law: A Useful Tool with Ethical Traps

BY CHRISTOPHER R. CASTELLANO, ESQ.

WHEN YOU GOOGLE (OR, FOR THAT MATTER, USE YOUR PREFERRED ARTIFICIAL INTELLIGENCE (AI) platform) “AI content,” you will find scores of outlets reporting on a stark finding that has developed in the last year, namely, society crossing the tipping point where more online content is generated by artificial intelligence than humans. This new reality should rightfully give you pause and make you consider its wider implications. Of course, cautionary tales and articles about the use of artificial intelligence may be as plentiful online as AI-generated content, but a polarized approach to AI is simply not logical. Artificial intelligence is here, and it is not going anywhere.

In the legal world, we need to appreciate the benefits of AI whilst recognizing the risks and learn where each of us can strike that balance because, whether we like it or not, our clients use it to draft contracts, organize financial documents, summarize communication, and generate case summaries before they speak with counsel. Similarly, as one attorney uses AI to help prepare deposition outlines, summarize discovery, compare agreements, and refine correspondence, their opposing party/counsel may use it to create polished but inaccurate narratives. With all litigants having access to tools capable of fabricating or altering texts, emails, photographs, audio recordings, financial summaries, and other evidence, the entire legal community must rise to the challenge of understanding AI, including its strengths and weaknesses.

As a family law attorney in Maryland, I have found that the question of AI has become an ever-present factor, from pro-se parties providing AI-drafted custody agreements to the nowfamiliar AI-formatted summary of a document. Therefore, the question is how Maryland family law practitioners can use AI responsibly while recognizing the evidentiary and ethical risks it poses. Used carefully, AI can be valuable for organizing, drafting, issue spotting, and preparation. Used carelessly, it can mislead the court, compromise client confidences, distort negotiations, and damage the lawyer’s credibility.

AI-Generated Material and the Evidence Problem

Family law litigation is especially susceptible to AI-related evidentiary problems because the evidence often consists of ordinary communications and personal records, including text messages, screenshots, social media posts, photographs, spreadsheets, and chronologies of alleged facts, just to name a few. The problem is not limited to dramatic deepfakes. The more common risk is subtler.

A party uses AI to summarize communications, but the summary omits context. A party creates a spreadsheet using AI-assisted categorization, but the categories are wrong. A litigant “reconstructs” text messages from memory and produces them in a screenshot-like format. A parent submits an AI-polished chronology that blends facts, assumptions, and advocacy. The danger is not merely that AI can create false evidence. The danger is that AI can make unreliable material look organized, neutral, and authoritative.

Md. Rule 5-901 requires authentication or identification as a condition precedent to admissibility. In Sublet v. State, 442 Md. 632 (2015), Maryland’s high court addressed authentication of social media evidence and emphasized that the trial judge must determine whether there is proof from which a reasonable juror could find that the evidence is what the proponent claims it to be. That framework is certainly relevant to AI-era family law practice.

Just as a screenshot is not self-proving because it looks like a text message, a social media post is not automatically authentic because it bears a party’s name. When analyzing potentially probative information for its evidentiary value, family lawyers should ask practical foundation questions to determine the document’s creator, source, and whether it was edited, enhanced, summarized, translated, or regenerated. One valuable piece of information that is all too familiar but remains no less important in the ‘AI-era’ is metadata. An understanding of metadata can help to gain insight into these questions and help verify the authenticity/ credibility of information.

Of course, the adage ‘trust but verify’ is never outdated. So why not pose a simple, practical intake question to the client about whether any document, image, recording, communication, timeline, or summary was created, edited, enhanced, translated, summarized, or organized using artificial intelligence?

AI in Negotiation

AI also has a growing role in family law negotiation. Used properly, it can help lawyers and clients organize issues before mediation or settlement discussions. It can create a checklist of unresolved custody provisions, compare versions of a proposed agreement, or even identify missing terms.

But AI should not be treated as a mediator, valuation expert, custody evaluator, financial analyst, final contract drafter, or judge. AI systems do not know the parties; they have no actual insight into your assigned judicial officer, your local county realities, credibility issues, or a client’s actual life needs. This is because of one immutable and simple fact—AI does not think, it merely generates. This is why AI may generate impressive settlement language that sounds sophisticated, but under scrutiny you may find it is vague, unenforceable, or inconsistent with Maryland practice.

Used carefully, AI can be valuable for organizing, drafting, issue spotting, and preparation. Used carelessly, it can mislead the court, compromise client confidences, distort negotiations, and damage the lawyer’s credibility.

This risk is acute when clients use AI before counsel is involved.

A client may arrive convinced that an AI-generated theory, custody proposal, or property division that AI said was “fair.” Or worse yet, they may input your proposals or work product into AI to determine ‘flaws.’ But one fact remains: AI is a tool, not an adjudicator.

The Maryland Warning: Mezu v. Mezu

Maryland family law now has its own cautionary AI case. In Chukwuemeka Mezu v. Kristen Mezu, No. 361, Sept. Term 2025, Slip Op. (filed Oct. 29, 2025), the Appellate Court of Maryland addressed a family law appeal involving fictitious AI-generated case citations. The court recognized that AI may be valuable in legal practice but warned that it must be used responsibly. It explained that AI “hallucinations” can result in fictitious cases or cases that do not support the propositions for which they are cited.

The lesson of Mezu is not simply “do not cite fake cases.” There are cautionary articles and seminars that abound with that precise thesis. The broader lesson is that AI does not dilute the lawyer’s personal responsibility for legal work submitted to your client or a court. The lawyer remains responsible for the pleading, research, citations, factual assertions, and

The duty of competence is central. Does this mean that lawyers now need to become AI engineers? No. But a lawyer should understand the tools they use well enough to recognize their risks. AI may hallucinate cases, misstate holdings, omit controlling authority, or present non-law as law. It may also omit, distort, or overstate important facts when summarizing discovery. So, like any tool, understanding its limits is critical.

Confidentiality is equally important. Family law files contain deeply sensitive information ranging from a child’s name to a party’s social security number, bank account numbers, settlement positions, abuse allegations, mental health history, substance abuse concerns, medical information, and all manner of privileged communications. Most AI platforms that you may be familiar with have data-retention policies, training practices, or human-review processes that are incompatible

The duty of competence is central. Does this mean that lawyers now need to become AI engineers? No. But a lawyer should understand the tools they use well enough to recognize their risks.

supervision of those who assisted in preparing the filing. Whether the work product is an appellate brief, a trial court motion, discovery responses, financial statements, proposed findings, child support worksheets, settlement agreements, or trial outlines. It is the lawyer’s signature at the bottom of the page and their responsibility to stand by the product.

Ethical Duties When Using Public AI Tools

There is no surprise that the commercially available AI tools raise several ethical concerns. American Bar Association (ABA) Formal Opinion 512 identifies professional responsibility issues implicated by generative AI, including competence, confidentiality, candor toward the tribunal, supervisory responsibilities, and fees.

with an attorney’s confidentiality obligations. Therefore, lawyers should be cautious before entering client facts, documents, or strategy into any AI system.

Candor, accuracy, and supervision are also implicated. Lawyers must review AI output to ensure that assertions made to a court are not false, including citations, legal analysis, facts, calculations, and arguments. Just as a lawyer cannot avoid responsibility by blaming a law clerk, paralegal, associate, or contract attorney for an inaccuracy in a filing, a lawyer cannot blame “AI” for the same. Firms large and small should have clear policies governing when AI may be used, what information may be entered, who must review the output, and what uses are prohibited.

Client fees present a final issue. If AI allows a lawyer to complete a task more efficiently, the billing must reflect the actual work performed,

Adopting a written AI protocol, identifying a clear firm-wide policy on the use of AI, and critically, understanding your desired platform can help transform a practice.

including time spent prompting, reviewing, correcting, and applying legal judgment, not an extrapolated charge if AI was not used. AI can help lawyers simplify tasks, but it remains the lawyer’s obligation to pass that time savings on to the client.

Proper Uses and Practical Guardrails

Despite these risks, AI can be used appropriately in the practice of law, including specifically, family law. The key is to use it as a tool. But it is critical to remember that, like any tool, AI is not a substitute for professional judgment.

Adopting a written AI protocol, identifying a clear firm-wide policy on the use of AI, and critically, understanding your desired platform can help transform a practice. Most responsible lawyers know that AI should not be used as the final source for legal research or drafting and that every case, statute, rule, quotation, and legal proposition must be checked against a reliable legal database or official source. Similarly, asking clients whether evidence has been created, altered, summarized, enhanced, or organized using AI is now standard practice.

Artificial intelligence will become a regular part of family law practice, if not the practice of law generally. The profession should not respond with panic, but with discipline. It is the attorneys who responsibly embrace AI that will help to forge a new and successful path forward, as opposed to those who reject AI, who risk being left behind. Yes, the world is flooded with AI talk, and the irony of this article is not lost on me. Nevertheless, recognize AI for its potential to enhance the practice of law, not a challenge to its existence. While it is true that many outlets report that AI content accounts for a majority of online content, those same outlets also note that human-generated content still garners higher engagement because the human mind simply cannot be replaced.

focuses his practice primarily on both contested and uncontested family law matters, including pre- and postnuptial agreements, separation agreements, divorce, marital property division, business valuations, child custody and visitation, spousal and child support, and modification.