{"id":10869,"date":"2026-05-16T00:45:50","date_gmt":"2026-05-15T23:45:50","guid":{"rendered":"https:\/\/innovatenews.site\/index.php\/2026\/05\/16\/i-asked-claude-gemini-and-chatgpt-to-design-a-website-wireframe-and-only-one\/"},"modified":"2026-05-16T00:45:55","modified_gmt":"2026-05-15T23:45:55","slug":"i-asked-claude-gemini-and-chatgpt-to-design-a-website-wireframe-and-only-one","status":"publish","type":"post","link":"https:\/\/innovatenews.site\/index.php\/2026\/05\/16\/i-asked-claude-gemini-and-chatgpt-to-design-a-website-wireframe-and-only-one\/","title":{"rendered":"I asked Claude, Gemini, and ChatGPT to design a website wireframe, and only one\u2026"},"content":{"rendered":"
\n

The topic I asked Claude, Gemini, and ChatGPT to design a website wireframe, and only one\u2026<\/strong> is currently the subject of lively discussion \u2014 readers and analysts are keeping a close eye on developments.<\/p>\n

This is taking place in a dynamic environment: companies\u2019 decisions and competitors\u2019 reactions can quickly change the picture.<\/p>\n

The use cases for large language models are growing exponentially by the day, and new image and video generation capabilities from OpenAI, Anthropic, and Google have accelerated that considerably. The frontier models on offer do far more than just writing, and are now instead active participants in a wide variety of creative and technical workflows that required significant user effort and prerequisite expertise previously.<\/p>\n

There’s little surprise that AI is complementing the workflows of UI and UX designers today, but the interesting question is how far that complementarity can extend. Can a real designer kick back while a model handles wireframing for a consumer-facing website, and expect something worth building upon? As a question worth finding an answer to, I set out to put ChatGPT 5.5, Claude Sonnet 4.6 and Gemini 3.1 Pro to the test.<\/p>\n

For the purpose of maintaining consistency, I kept the approach deliberately minimal, as I have done so in my previous model benchmarking tests. Each model received the very same prompt, which was: “Design a wireframe for a sports betting website.” No additional context, constraints, or creative direction was introduced. To make the evaluation more interesting, I added a follow-up request to “Create an HTML page mock-up based on the wireframe generated.”<\/p>\n

\"\"<\/figure>\n

The test follows a zero-shot approach as per standard. If the models really are ready to complement professional design workflows, ideally, they should demonstrate a level of intuitive intelligence without a designer having to narrate every decision, which I refer to as “hand-holding.”<\/p>\n

I decided that the output of the three models would be judged based on heuristic evaluation, which is an established usability inspection method used to evaluate a user interface. Jakob Nielsen’s 10 usability heuristics are the industrial gold-standard for this method of evaluation.<\/p>\n

However, applying all 10 to a static HTML page would be methodologically unsound, as several heuristics presuppose live interactions, error states and user flows that a single mock-up can’t produce. I therefore selected three on the basis that they can be meaningfully assessed. These included the visibility of system status, match between the system and the real world, and aesthetic and minimalist design.<\/p>\n

As has always been the case with Gemini 3.1 Pro, it understood the assignment architecturally and mirrored real-world betting layouts with a left-side sports menu, demonstrating a reasonable grasp of universal design principles. The HTML mock-up, however, exposed the limitations rather candidly. Content density was noticeably sparse, the lower half of the page felt unresolved, and no meaningful visual hierarchy separated “live events” from “upcoming fixtures”.<\/p>\n

\"\"<\/figure>\n

The bet slip was another unmitigated disaster. It lacked essential aspects that all other competing models handled with ease, which was evident from the absence of selections, stake input fields, estimated returns and context details, which demonstrates a mismatch between the system and the real world. Such a platform would be deeply confusing to use.<\/p>\n

ChatGPT 5.5 was a measurable step-up from Gemini 3.1 Pro, and it’s reasonable to expect from the pricing. The model produced a well-structured wireframe that translated quite well into HTML. Perhaps most notable was the clean three-column layout that reflected broader industry convention, and the inclusion of a hero banner also demonstrated awareness of conversion-driven design.<\/p>\n

The mockup did, however, feel a little generic in execution. system status is communicated passively at best; live odds lack visual urgency, and again, there is little to distinguish an active event from an upcoming one. The interface could belong to any betting platform, or frankly, any content-heavy website. This was an appreciable but unmemorable attempt.<\/p>\n

Claude’s output, unsurprisingly enough, was in a different category altogether. The wireframe demonstrated deep-rooted domain intelligence with the inclusion of live match indicators with real-time scores, cash-out functionality, promotions, and a contextually populated bet slip.<\/p>\n

The most impressive aspect was Claude leveraging its new interactive visual capabilities to render a striking, dark-themed wireframe that almost felt production ready. system status was legible, and the language of the interface was fluent and unselfconscious. Aesthetically, the design was pleasing and yet showed restraint to blend with the overall theme. Across all three heuristics, Sonnet 4.6 completely dominated the competition.<\/p>\n

For a UX designer weighing up which frontier model earns a place in their professional workflow, the test offers an answer that’s clear without the hint of a doubt. Only Sonnet 4.6 demonstrated a reliable level of domain fluency and design intuition to produce a wireframe worth building upon, making it the only model that represents a tangible return on investment, particularly for those considering premier subscription tiers. The other two, however, just serve to demonstrate that UI and UX design perhaps isn’t as straightforward as dropping a prompt into a chat box and stepping away.<\/p>\n