There’s no doubt that AI is going to change divorces. I’ve been a family lawyer for a while and am a bit of a technology nerd. Over the last year, I’ve actually used AI like Caseway and LexisNexis+ in real divorce cases. I figured I’d share some firsthand impressions. This is what actually happened in the trenches of my divorce practice.

Where AI has been a lifesaver

Legal research in divorce cases can be a slog. Before these kinds of software, I’d spend late nights digging through Lexis or Westlaw for that one case. Let’s say that a spouse hid cryptocurrency assets or something oddly specific. Now, with things like Caseway (like a Chatgpt trained only on case law), I can type a plain-English question and get relevant court cases in seconds.

For example, I had a case about dividing airline miles in a divorce, which is not exactly a common issue. I asked the machine, and boom! A couple of cases (and even a law review article) surfaced that dealt with splitting travel points. I may never have found those with manual keyword searches. It saved me hours. My client was impressed that I found a precedent so fast (I didn’t mention the AI; I just took the compliment).

The drafting help is pretty neat, too. LexisNexis+ has this AI assistant that can summarize cases or even throw together a first draft of a memo. I had to write a brief on a convoluted child custody jurisdiction issue. There were two states, both claiming authority, which was a real headache.

I let the AI give me a draft outline. It was surprisingly decent. It laid out the relevant factors and even cited the Uniform Child Custody Jurisdiction Act sections. That gave me a solid starting point. It’s like having a super-fast paralegal who’s read every case ever.

I still had to finesse the language and add the emotional appeal, since no AI can truly capture the “think of the children” vibe needed in custody fights. But structurally, it got me 70% there in a few minutes.

Quick research payoff

In one divorce case, my client was convinced her ex was underreporting income. There’s a newer law here about gig economy income in support calculations. In recent cases, I wasn’t sure how courts were handling Uber/Lyft earnings. Rather than manually search, I asked the AI assistant.

It quickly spit out a summary of a 2022 case where a judge imputed income based on gig earnings. That was gold for us. I read the full case to ensure it said what the AI claimed (it did), and we used it. The judge was a bit surprised I had a new case at my fingertips. We secured a better support order partly because of that little AI assist.

Where it can go wrong or fall short

It’s not all roses. These kinds of AI software have quirks and limitations. For one, they’re only as good as the info you feed and the questions you ask. I learned this the hard way. Early on, I tried Caseway for a relocation case (one parent wanted to move states with the kid). I hastily typed something like “Can a mother move the child out of state after divorce if the father opposes?”. It gave me a case that looked on-point.

Thank God I read it closely. It was a Canadian case! I wasn’t specific enough, and Caseway’s database at the time was Canada-focused. I’d have face-planted in court if I had just skimmed the summary and cited it. I had to refine the query (“California relocation child custody recent cases”) to get valuable results. So, lesson learned. You still need research skills even with fancy AI. Garbage in, garbage out.

AI also doesn’t know your judge or the “unwritten rules.” In family law, that stuff is huge. Like, one judge in our county always favours 50/50 custody unless you show extreme circumstances. No research AI will tell you that. It’s not in any case or statute; it’s just her style. An experienced attorney or local will know, but AI won’t.

Same with how a judge reacts to specific arguments or evidence. AI might suggest a perfectly legal strategy that theoretically should work, but you’re sunk if it triggers a judge’s pet peeve. I’ve had AI suggest raising a minor point of law that was technically valid. But I knew the judge wouldn’t care, and it’d just annoy him. That insight comes from humans talking to humans.

AI’s lack of context and emotional intelligence

Divorce isn’t just law; it’s raw emotion and human messiness. AI is tone-deaf to that. I tried using AI to help draft a client communication once, basically explaining why our settlement offer was fair. The draft it gave was super logical and well-structured, but it came off cold and kind of patronizing.

A line like “According to subsection X of the code, your position lacks legal merit” isn’t how you tell a distraught parent they might not win on an issue. The AI didn’t get the need for empathy and diplomacy.

I scrapped most of it and rewrote it in a more humane tone. Clients don’t want a robot; they want to feel heard. So I use AI for the analytic stuff, but I’d never let it near sensitive client conversations or judgment calls about how hard to push. It has no gut feeling or moral compass.

Hallucinations and mistakes

We’ve all heard about the lawyer who cited fake cases because ChatGPT just made them up. In my practice, I haven’t had Caseway or Lexis’s AI concoct completely fake citations (they promise they use real databases only). But I have seen smaller errors. One time, Lexis+ AI summarized a case and misstated the holding slightly.

It said the judge “denied” something the judge criticized but still allowed. I would’ve misrepresented the precedent if I hadn’t caught that by reading the actual case. It wasn’t a full hallucination, more like a distortion. Still dangerous. Now I double-check any case the AI finds or summarizes. As one legal tech buddy told me, “trust but verify.” These are words to live by.

Another glitch is jurisdiction differences. Family law is super state-specific (and country-specific). I asked the AI to draft arguments for dividing marital property in a common-law state. The draft kept mentioning “community property.” These were true presumptions in some places (like California) but not where I practice. The AI wasn’t wrong generally; it just didn’t grasp that those rules don’t apply everywhere.

It had mixed principles from different jurisdictions. Imagine a non-lawyer seeing that draft; they’d be completely confused or think they’re entitled to 50% of everything because the “AI lawyer” said so, even if their state doesn’t do 50/50 splits. This is why AI is not a DIY lawyer for the public.

If you don’t know the law, you might not realize the tool gives you out-of-context info. I’ve had to correct people on Reddit who confidently quoted “the law” per Chatgpt, but it was entirely the law of another state.

Almost got burned

I’ll share a specific close call. A colleague was curious and used GPT-4 (not through Lexis, just the vanilla Chatgpt) to help draft a motion for a discovery issue in a divorce. He knows tech, but he got a little overconfident. The draft that came back looked legit, with citations and all. He almost filed it as-is.

Something felt off to him (maybe it was too clean), so he ran the cites through Westlaw. Sure enough, two cases were completely made up. This included the quotes, the docket numbers, everything. If he had filed that and the judge or opposing counsel caught it, his reputation could’ve been toast.

He rewrote it using traditional research, but that scare showed us both: you can’t just blindly trust an AI. Even the legal-specific ones, like Caseway, I still verify every case. I treat them as a super-smart intern who sometimes lies or bluffs when it doesn’t know the answer. You gotta check the work.

Not replacing real lawyering

These AI software platforms are like a power drill for a carpenter. They’ll do the tedious, heavy lifting tasks faster, but won’t design the house for you. In my divorce cases, I still have to exercise legal judgment. Examples of this are what strategy to pursue, how to negotiate with the opposing party, how to handle my client’s emotional state, and how to present things in court.

AI doesn’t do any of that. It won’t magically understand that my client wants to fight out of principle, or that the opposing lawyer has a massive ego I must massage in negotiations. It just provides data and drafts.

AI Is Going To Replace Divorce Attorneys?

I also consider the latest human nuances. For instance, our legislature passed a new custody law last year in response to a high-profile case. No cases are interpreting it yet, but I know the intent from legislative history and local news. AI doesn’t have that kind of context (at least until those cases become part of the dataset).

Similarly, the local court culture is shifting on spousal support due to the economy. You only pick that up from being in the field, talking to other lawyers and judges. AI reads text; it doesn’t attend bar mixers or judges’ lunches (ha!).

AI in divorce law is a game-changer for efficiency. It helps me find relevant laws faster and draft routine documents more quickly. It’s cut my research time significantly, which can translate to lower bills for my clients (or just me getting to sleep a bit more). But it’s not a game-changer for judgment, strategy, or empathy. Those remain very human.

If you’re dealing with a divorce (as a lawyer or on your own), feel free to leverage these tools for information gathering. Just don’t mistake them for a lawyer’s brain. They don’t understand your unique situation’s subtleties or the human heartache involved.

Using AI as an assistant

In my experience, the best results come when I use AI as an assistant, not an oracle. I’ll take its suggestions, compare them with my own research and experience, and then make the call. Thanks to AI, it’s pretty awesome to have a precedent from 2023 at my fingertips. This is especially true when opposing counsel cites stuff from 1995 because he didn’t know it existed.

In that sense, it can give you an edge. But I’ll never go into court or a settlement relying solely on an AI’s answer.

Hope this gives some real-world perspective. If you’re curious about using AI in your case or practice, go for it. Just keep your legal common sense turned on at all times. And if something seems off or too good to be true (like a perfect case quote that just happens to exactly support you), double-check it. Trust me on that one.

Source: I’m a divorce lawyer who’s actually used this kind of software, not just read about it. (Also, there was a great discussion on Real Talk with Ryan Jespersen about AI shaking up divorce law if you’re interested.)