Ask an AI to build a working web page instead of just chatting — how long is the wait, and is it worth it?

This asks how long it takes an AI model to build a complete, working web page as its answer, rather than replying in chat text. The short answer is that the collection doesn't give a clear number. None of the notes here measure how many seconds or minutes a model needs to generate a page. What the collection does cover is the related question: whether that page is worth waiting for, and how often it comes out right.

The case for waiting is strong. When people compared full web pages generated by an LLM with ordinary formatted chat replies, they preferred the web pages 83% of the time. The generated pages matched pages built by human experts about half the time, and only the newest models could do this at all Do full web pages beat markdown chat for LLM responses?. A related line of work found that interfaces built for a specific task, such as dashboards, small tools and animations, beat plain text in more than 70% of cases. The biggest gains came on tasks with a lot of structured information, where a laid-out interface makes the reader do less work Do generated interfaces outperform text-based chat for most tasks?. So the useful question may be less about how long generation takes and more about how much of the reader's time the page saves afterward.

The less obvious finding is that a finished page isn't always a correct one. A benchmark of five generative UI tools found that about a quarter of the design decisions a tool claimed to have made never showed up in the page it built. For functional requirements, meaning the parts that are supposed to actually work, that gap rose to 34% Do generative UI tools actually implement their stated design rationales?. In practice, the real end-to-end time includes checking the result and asking for fixes, not just the first generation. A study of AI-generated analysis interfaces points the same way. The widgets made results clearer, but they were harder to change partway through a task, so revisions cost extra prompting Do generated analysis UIs really work better than chat?.

The closest thing to a timing result in the collection measures people, not models. In a randomized trial with 100 designers and product managers, those given Figma Make, an AI design tool, finished structured design tasks about 20% faster, and product managers gained more than designers Does Figma Make speed up design task completion?. That suggests the time that matters in practice is the whole loop of prompt, generate, review and revise, and AI shortens that loop without removing it. If you want model-side speed numbers, they would have to come from sources outside this collection.