At first glance, the 2025 case of Dewald v. MassMutual feels more like satire than a real appellate decision. A pro se litigant attempted to present oral argument through an AI-generated avatar video instead of addressing the court himself. But beneath the awkward spectacle was a serious lesson about generative AI and the limits of trying to outsource human judgment.

Within seconds, the panel stopped the presentation. The judges were not amused. They questioned whether the court was being misled and made clear that if the litigant wanted to be heard, he needed to stand up and argue his case himself.

It is easy to laugh at the optics of a “virtual lawyer” making an appearance in a real courtroom. But underneath the awkwardness is the more serious issue of the limits of generative AI, especially in the hands of unrepresented parties.

Pro se litigants already face an uphill climb. They are navigating complex procedural rules, substantive law, and courtroom dynamics without the benefit of training or experience. Generative AI can feel like a lifeline in that situation. It offers quick answers, polished language, and the illusion of expertise. But that is exactly where the danger lies.

AI tools are very good at producing something that sounds right. They are far less reliable at being right in the context that matters. Legal arguments are not just about stringing together plausible sentences. They require judgment, nuance, and the ability to respond in real time to questions from the bench. An AI-generated script cannot do that. It cannot adjust when a judge interrupts. It cannot clarify a point that lands poorly. It cannot read the room.

That is what the court in Dewald seemed to be pushing back against. Oral argument is not a prerecorded performance. It is a conversation. Judges use it to test arguments, probe weaknesses, and sometimes even guide a party toward a more effective framing of the issue. By outsourcing that moment to an avatar, the litigant effectively opted out of the most valuable part of the process.

There is also a credibility problem. Courts depend on candor and authenticity. When a litigant presents an AI-generated figure without clearly framing it for what it is, it risks looking like an attempt to mislead, even if that was not the intent. That alone can undermine whatever substantive points the party is trying to make.

None of this means that AI has no place in legal work. It can be a useful tool for organizing thoughts, summarizing information, or even generating a first draft of an argument. But Dewald is a reminder that there is a line between assistance and substitution. When AI crosses into replacing the human advocate, it stops being helpful and starts being harmful.

For lawyers, the lesson is relatively straightforward. Use AI, but do not surrender your role to it. Your value is not just in producing words. It is in exercising judgment, adapting in the moment, and maintaining credibility with the court.

For pro se litigants, the stakes are even higher. AI can support your preparation, but it cannot stand in for your voice. If anything, relying too heavily on it may widen the very gap you are trying to close.Dewald v. MassMutual may be remembered for its unusual facts, but its message is clear. In a system built on human judgment and accountability, there are still moments when you have to stand up and speak for yourself.