I Built a Lawyer Game Where the NPCs Can Actually Argue Back

I finally built a lawyer game with AI.

It’s called Legal Arena, and it started with a question I’ve had for a long time: what would it be like to have AI-powered NPCs in a game?

In traditional games, most NPCs are scripted. Those scripts are often called “AI,” and they can be incredibly detailed, but the characters still behave in a mostly deterministic way. They respond with lines someone wrote in advance. The player chooses from a list of dialogue options, and the NPC follows the corresponding branch.

Modern AI makes a different kind of interaction possible.

You can talk to a character in freeform text. The character can react to what you actually said, reveal information based on the conversation, and change the direction of the game. Instead of selecting the closest available dialogue option, you can make your own argument.

I wanted to build a game around that idea. I also wanted to argue with a computer and see whether I could genuinely get a point across.

A legal game felt like the most logical place to start.

Why a Lawyer Game?

Law is fundamentally built around language, facts, persuasion, and competing interpretations.

A lawyer has to interview a client, figure out what happened, identify useful evidence, understand weaknesses in the case, negotiate with the other side, and eventually make an argument. Those actions map naturally to conversations with AI characters.

That became the foundation of Legal Arena: a game where you play as a lawyer and interact with clients, opposing counsel, and judges through open-ended dialogue.

The NPCs aren’t simply waiting for you to click the correct pre-written response. You have to ask questions, uncover information, choose a strategy, and make your case in your own words.

Creating a Case

A game begins by choosing a genre of dispute.

There are around 11 genres, including:

• Rental
• Marital
• Criminal
• Administrative
• Employment

You also choose the toughness level of the case.

Based on those choices, the game generates two sides of a dispute: a plaintiff’s story and a defendant’s story. You are then randomly assigned to represent either the plaintiff or the defendant.

That random assignment matters because you don’t always get the stronger or more sympathetic side. You have to work with the client and facts you are given.

From there, the game has three stages:

1. Intake
2. Settlement
3. Courtroom

Settlement and courtroom are alternative paths. Depending on what you learn during intake, you can try to settle the dispute or take the case to court.

Interviewing Your Client

The intake stage is where you first meet and interview your AI client.

You have to find out what happened by asking questions. As the client reveals information, the game populates something called a fact sheet.

The fact sheet is the source of truth for the case. It contains condensed information such as:

• Proven facts
• Corroborated facts
• Disputed facts
• The case timeline
• The theory of the case
• The relief being requested from the court

This is the material you will later rely on if you decide to fight the case in court.

The client may not immediately tell you everything. You have to conduct the interview and discover the facts through conversation. What you ask—and how thoroughly you ask it—affects how much of the case you understand.

Once the fact sheet has been populated, you should have a rough idea of your chances in court.

Sometimes the case looks strong. Sometimes it becomes clear that your client is missing important evidence. You may also discover facts that weaken your own position.

At that point, pushing ahead to court may not be the smartest move.

Settlement Is a Three-Way Negotiation

If the evidence isn’t strong enough, or if going to court seems too risky, you can attempt to settle with the opposing party.

I wanted settlement to be more than a simple exchange between two lawyers. In Legal Arena, it is effectively a negotiation among three parties:

• You
• Your client
• Opposing counsel

Your own client may not automatically agree with the settlement you think is reasonable. At the same time, opposing counsel has to be persuaded that the proposed terms are worth accepting.

The quality and likelihood of a settlement depend partly on the mood of both your client and the opposing counsel.

You can talk to your client and try to improve their mood or make them more receptive to a compromise. You can also make proposals to the opposing side and try to move the negotiation forward.

The better the mood on both sides, the greater the chance that the terms of a settlement will be accepted.

This creates a different kind of challenge from the courtroom. You aren’t just proving that your position is correct. You are trying to manage expectations, maintain cooperation, and find terms that multiple people will accept.

Taking the Case to Court

If the case doesn’t settle, you can fight it in court.

Here, you have to present your argument. You need to use the information uncovered during intake, refer to the available evidence, and make the strongest case possible for your client.

The opposing counsel is also AI-generated and presents an argument for the other side. An AI judge then evaluates the quality of both arguments.

There is no single dialogue button marked “good legal argument.” You have to construct the argument yourself.

That is the part of the game I was most interested in building: a situation where the player can’t progress by guessing which scripted response the designer intended. The player has to understand the case and communicate a convincing position through freeform text.

You have to be a good lawyer—or at least make the best attempt you can with the facts available.

Building a Legal Career

Legal Arena also has a bar association where you can see the other human players in the game.

You can view their records, follow how they have progressed through their careers, and see which kinds of disputes they are developing expertise in.

You can also challenge another player to a player-versus-player match.

PvP follows a structure similar to the regular game, but it is played asynchronously. Each human player separately interviews their own AI client. Once both sides reach court—or decide to negotiate a settlement—they begin interacting across the dispute.

In court, both players make their arguments. During settlement, each player still has to communicate with their own client while negotiating with the other side.

The AI characters remain part of the case, but the opposing strategy now comes from another person.

NPCs That Respond to What You Actually Say

The central idea behind Legal Arena is simple: NPCs should be able to respond to the player’s actual words.

The client can reveal information based on the questions you ask. Opposing counsel can react to the proposals you make. A judge can evaluate the argument you present. None of this requires the player to select from a list of pre-written dialogue choices.

That makes Legal Arena one of the first games where I’ve been able to explore the kind of AI-driven NPC gameplay I had imagined: conversations that aren’t just decoration, but the mechanism through which the game moves forward.

The Case Is Open

Legal Arena combines investigation, negotiation, evidence, and freeform argument into a game about practicing law. Sometimes the right move is to fight. Sometimes it is to settle. Either way, the result depends on what you discover and what you manage to say.

The game is available at legalarena.app.