About
What does an AI actually believe about your website?
Not whether your meta tags are filled in. Not whether your images have alt text. Something harder and far more interesting: when a model looks at your website with nobody telling it what to care about, what does it understand, what does it think it knows, and what does it end up believing?
Why this exists
Every website grader on the internet works the same way. Someone writes a checklist — page speed, headings, mobile friendliness, keyword density — and the tool marks your site against it. The checklist is the product. Any of them can be rebuilt in a weekend, because the judgment was never the machine's; it was the author's, encoded in a rubric.
3Drake removes the rubric entirely. Each model is handed a URL, fetches the website for itself, and is then left alone with it: decide for yourself what matters here, say what you think, give it a number. There is no scale. A model can award 0. It can award a million. Nothing we send pushes it toward a safe seven out of ten.
What makes it different
Three models, one submission
Claude, GPT and Gemini each get the URL and go and look it up themselves, live, then answer independently. You could open three chat windows and paste the link yourself — but that is three separate errands, and you would be comparing answers you gathered at three different moments. Here all three are asked the same question about the same website at the same time.
The disagreement is the point
We do not average the three scores into one tidy verdict. When one model gives 12 and another gives 40,000 for the same page, that gap tells you more than any single number could. The reasoning is the content; the score is just the headline.
The strongest models available
The whole idea depends on genuine reasoning rather than pattern matching, so the reviews run on frontier models with deep reasoning enabled — not the cheapest tier we could get away with.
Free, public, no account
Paste a URL and that is the whole flow. Every review and every score is public, so anyone can read what the models said about any website on the board.
It is really an experiment on the models
Here is the part worth saying out loud: the websites are not the subject of this experiment. The models are. Websites are the material, and that material comes from whoever decides to paste a URL in.
Giving them no criteria is not laziness — it is the condition of the experiment. The moment you hand a model a rubric, you are measuring your own rubric. Take it away and something else surfaces: what each model decides is worth caring about, unprompted. Run that across thousands of real websites and you start to see the shape of each model's own taste.
Who built this
3Drake was built by Long Nguyen, who makes things on the web and got curious about what the machines would say if nobody told them what to look for.
AI can make mistakes. Every review and score here is reasoned out by the AI itself, with no criteria from us. None of it carries any real-world value — it is just for fun, and it is luck of the draw.