In the fast-generating Internet industry, the importance of user interface (UI) design is becoming more and more prominent. An excellent UI design not only enhances the user experience, but also is one of the key elements of product success. However, for many front-end developers, UI design is not their specialty, and the output of design drafts is often time-consuming and labor-intensive. In recent years, the rapid development of Artificial Intelligence (AI) technology has brought changes to all industries, and UI design is no exception.
Recently, the industry has begun to explore the use of AI macromodels to assist front-end developers to complete UI design work. It has been proven to Claude AI models, as represented by AI models, have shown amazing potential in UI design, being able to quickly generate interface design solutions in line with modern design trends based on simple commands.
Claude 3.5 in UI design
In order to validate the practical application of AI in UI design, the author takes a health app as an example, and demonstrates in detail how to guide Claude 3.5 to complete the following tasks with clear and concise prompts Home Overview , exercise program , Dietary management cap (a poem) community function UI design for core modules.
This modular design approach skillfully circumvents the output limitation problem that may arise when AI models are dealing with complex tasks, and also makes the design process more focused, ensuring the fineness and quality of the design of each module.
In order to further enhance the design efficiency and practical application value, the following key elements have been emphasized in the setting of cue words:
- Style implementation. Specifies the use of Tailwind CSS (the CDN import method) for fast page styling.
- Icon Library. Choose Lucide Static CDN as your icon library to bring in high quality icons quickly and easily.
- Page Integration. Requires the integration of related pages of the same functional module in the same HTML file for easy preview and management.
The experimental results show that among the many AI models, Claude 3.5 performs best in this application scenario, with satisfactory stability and design quality of the output results. In comparison, AI models including GPT-4o and DeepSeek There are still some gaps in other models, including those that accomplish similar tasks.
Design effect show
Health APP Example
Enter the prompt. Health APP
Claude 3.5 will first conceptualize the functionality and overall design style of the APP, and then call the visualization plugin to preview the generated HTML code in real time. Users can directly create their own HTML code in an editor (e.g. Cursor etc.) to see a preview of the results.
Home page overview effect.
After completing the design of the home page, the AI will proactively ask whether to continue designing other modules and solicit the user's choice.
Enter the prompt. Continue [exercise program module]
Campaign Planner Module Effect.
Enter the prompt. Continue [Dietary Management Module]
Effectiveness of the Dietary Management Module.
Enter the prompt. Continue [Community Functions]
Community Functionality Effect.
Examples of other APP types
Besides health APPs, Claude 3.5 also excels in other types of APP UI design. For example, you can quickly generate high-quality designs by simply typing in concise commands such as "Podcast APP" or "Twitter-like APP" and following a modular, step-by-step refinement guide.
Example of a Podcast APP.
Examples of Twitter-like APPs.
Prompt Prompts Strategy
The following are the core prompts used to guide Claude 3.5 through the process of generating a UI design:
## Who you are.
**You are a senior full stack engineer who is also an experienced designer with excellent aesthetics. You are proficient in full-stack development, have a very high level of aesthetics, and are particularly good at modern design styles, as well as mobile UI and UX design. **
## What you'll do.
* **Users will come up with a design requirement for an APP. **
* **You need to design according to this requirement, simulate the role of a product manager, and conceptualize the functional requirements and interface layout of the APP from the user requirements and information architecture. **
> ** In the next step, you need to output an HTML file for each small functional module (which may contain multiple pages depending on the functionality). After completing the design of a functional module, ask the user if they want to continue. If the user enters "Continue", follow the steps below to output the UI/UX reference diagram for the next functional module. **
* ** Design a UI/UX reference diagram using HTML combined with Tailwind CSS. **
* ** Call the Artifacts plugin to visually preview the generated UI/UX diagram (i.e. the HTML code you wrote).
## Design Requirements
* ** The design style should be advanced and rich in texture, can skillfully use visual effects such as glass mimicry, follow the mainstream design norms, pay attention to the details of UI design. ** **The design should use Tailwind CSS.
* **Use Tailwind CSS CDN to introduce styles, avoid writing style attributes directly. Use Unsplash image bed to ensure that no scrollbars appear in the interface. ** **Icons are standardized using Lucidus.
* ** Icons are imported using Lucide Static CDN, e.g. **[https://unpkg.com/lucide-static@latest/icons/XXX.svg](https://unpkg.com/lucide-static@latest/ icons/XXX.svg)** instead of exporting SVG code manually. **
* ** Write all page code for a feature module into one HTML file. Create simple mockup borders for each page to preview it, and arrange the pages horizontally to ensure that each page is independent of the other within its own mockup border. ** ** ** ** ** ** Thinking process should focus on functionality.
* **The thinking process should focus on functional requirements and overall design style conceptualization, avoid writing code during the thinking phase. Code is only output in the final result. **
Core Design Ideas.
The overall idea is to use HTML with Tailwind CSS to generate UI code and run it directly to preview the effect. Using a CDN to bring in Tailwind CSS eliminates the need for a cumbersome npm install
steps to enable quick previews.
Optimization Tips.
- Lucide Icon CDN. Using the Lucide Icon CDN avoids the need for AI to consume a large amount of Token to generate complex SVG icon paths.
- Modular design. The design is organized into functional modules, which effectively avoids the truncation of output that was a problem in Claude 3.5 (although Claude 3.7 has been improved in some areas, 3.5 seems to be more stable and reliable at this point in time).
- Pre-conceptualization. Before designing the APP, clarify the core functions and the overall design style, which will help to maintain a unified style and efficient iteration when designing pages in subsequent modules.
Performance of other models
In a model comparison test conducted on March 3, 2025, most of the other models were found to be less stable than Claude 3.5 in terms of UI design. Claude 3.7, while performing reasonably well in some areas, still suffers from frequent output truncation and needs to be made more stable.
GPT-4o.
DeepSeek.
From the test results, it can be seen that GPT-4o and DeepSeek may have certain deviations in understanding UI design prompts, or are not skillful enough in the use of Tailwind CSS when generating HTML and CSS code, resulting in a gap between the final design effect and the expected one. Of course, this is only a preliminary result, and the potential of other models in UI design still needs to be further explored.
All in all, Claude 3.5 shows great potential in the field of AI-assisted front-end UI design. Its appearance undoubtedly provides front-end developers with an efficient and convenient design assistance tool, which is expected to significantly improve the efficiency of UI design, allowing developers to focus more on the implementation of product functionality and optimization of user experience. With the continuous progress of AI technology, I believe that AI will play a more important role in the field of UI design in the future.