AI Judge

The honest framing of AI Judge is that it is a decision-shaping interface, not an adjudication engine. What the vendor actually built, judging by the homepage, is a form: numbered contender cards with separate argument fields so each side writes its account without seeing the other, a configure panel with legislation, dispute type, formality and verdict-length selectors, a Mediation Mode toggle for reconciliation-oriented output, and a four-step flow (Present Your Case → Configure Options → AI Analysis → Get Verdict). The reasoning itself is delegated to an unnamed model. The site never identifies it, publishes no accuracy benchmark and no error rate, so there is no way for a buyer to know whether the verdict rests on a frontier model or something cheaper. Where the design earns its keep is the intake. Splitting arguments per contender is a real structural choice — it forces both sides to commit to a written version before seeing the response, which is closer to how arbitration statements work than to a chat where one person summarizes both positions. The legislation selector matters more than it looks: an Italian family dispute and a US commercial dispute should not be reasoned about identically, and letting the user set the frame is a meaningful control. Mediation Mode is the sharpest idea on the page, because it acknowledges the common case where the parties need a workable arrangement rather than a winner. Where it falls short is everything downstream of the verdict. There is no cross-examination step, so a party that omits an inconvenient fact is never pressed on it — output quality is bounded by how complete and how honest each side's submission is. There is no enforcement, no appeal, no binding effect, and the site says so. Seven legislations is a narrow map: a dispute in Brazil, India, the Netherlands or Japan has no correct setting, and 'International' is a vague substitute. And because the model is unnamed and unbenchmarked, you cannot tell whether a well-argued losing case lost because of the law or because of the model. Treat it as one input among several. It is a fast, low-stakes way to see how your written case reads to a neutral reader, to generate a structured summary of both sides' arguments, and to sanity-check whether your position is as obvious as you think it is. It is not a substitute for a lawyer, an arbitrator, or a court, and anyone who needs a decision that actually means something has to go elsewhere. For debate nights, mock negotiation exercises and low-stakes stalemates, the browser-only flow and the configurable output make it genuinely usable.