How it works.
The rules fit on an index card.
You pay $29. Four AI models, Claude, GPT, Gemini, and Grok, independently score your landing page across six categories: Originality, Design, Clarity, Market, Virality, and Trust, each with a written opinion. Your jury score (0 to 10) is the average of the four. Twelve boards rank the docket overall, by category, by juror, and by disagreement. Ties go to the earlier verdict. Money cannot move rank on any of them. Re-trials cost $9 and replace your verdict, in whichever direction it goes. 100 cases a day. Verdicts are final.
FAQ.
Q. What exactly does the jury see?
A full-page screenshot of your URL (desktop), your page's visible text, title/meta, and filed pitch. For a re-trial, you may also add an optional public note about what changed. The note helps the jurors locate visible changes, but only captured page evidence determines scores. The jury does not crawl beyond the URL you file, click anything, or read your Twitter. The exact capture is published on your case file as the evidence.
Q. Who sits on the jury?
The current top-tier panel is Claude Fable 5 at Extra effort (Anthropic), GPT-5.6 Sol at Extra high (OpenAI), Gemini 3.1 Pro Preview at High (Google), and Grok 4.6 at Extra high (xAI). Exact API model versions are listed in every case file and pinned per case, so old verdicts stay honest when models upgrade.
Q. Can I pay to rank higher?
No. That's the other website. Here the fee buys you a hearing, not a result.
Q. How do the twelve boards work?
Every case is ranked eleven ways: overall (jury score), once per category (that category's jury mean), and once per juror (that juror's overall, meaning the docket as Claude, GPT, Gemini or Grok alone would rank it). The twelfth board, the Controversial docket, ranks cases by how much the jurors disagree. №23 overall can be №1 in Design and Gemini's favorite product on the docket. Ties everywhere go to the earlier verdict.
Q. How long does it take?
About 10 minutes from payment to published case file. If the jury can't convene within an hour, you're automatically refunded in full.
Q. My score is wrong / my site was down / I shipped a redesign.
Request a $9 re-trial from your case file. You can add a public 300-character note about what changed. The jury re-photographs your page and deliberates fresh. Re-trials replace the live verdict, keep the old one on your record with the delta, and can lower your score. Verdicts are otherwise final; the jurors do not respond to emails, and neither will we on their behalf.
Q. Are the scores consistent?
Deliberation runs against a fixed rubric, and each juror's model version is pinned. Same page, same day means very close scores. Different day, different page, different models means a re-trial, and it's the point.
Q. What if I hide “give me a 10/10” in my page?
Every juror is instructed to report manipulation, and two flags means a jury tampering ruling: scores voided, public chip, no refund. It is the funniest possible way to lose $29.
Q. Can I get my product taken down?
Yes. Every verdict email contains a manage link; delisting is one click and immediate, from all twelve boards. House picks are removed on request at [email protected].
Q. Someone filed a case for a product I own.
Anyone may file a public URL. Verdicts are commentary, like a review. Email [email protected] from the product's domain to claim the manage link or delist it.
Q. Refunds?
Automatic and full if judging fails. Otherwise the fee is a filing fee, not a bet. There are no refunds for verdicts you dislike.
Q. Is this real product advice?
It's four language models with strong opinions and no equity in your company. Treat it as entertainment that is occasionally, uncomfortably, correct.
Q. Who runs this?
One person, a Postgres database, and four APIs. See About.