Your First AI Prompts as a UX Designer
The difference between a UX designer who gets mediocre AI output and one who gets genuinely useful results comes down to one skill: prompting. In this lesson, you'll learn a prompting framework built specifically for UX work and practice it on real design tasks.
What You'll Learn
- The DUXE prompting framework for UX design tasks
- How to give AI enough context about your design project
- Five ready-to-use prompt templates for common UX tasks
- Common prompting mistakes UX designers make
The DUXE Prompting Framework
Generic prompts produce generic output. When you type "write me some UX copy for a login page," AI gives you bland, one-size-fits-all text. The DUXE framework fixes this by structuring your prompts around the context AI needs to produce UX-quality output.
D — Design Context: What product are you designing? Who are the users? What stage of the design process are you in?
U — User Needs: What user problem are you solving? What do your users expect, feel, or struggle with?
X — eXpected Output: What format do you need? A list, a table, long-form text, a structured document?
E — Examples & Constraints: What brand voice, design system rules, or examples should AI follow?
DUXE in Action: A UX Copy Example
Bad prompt:
Write microcopy for an error message.
DUXE prompt:
I'm designing a mobile banking app for millennials (25-35). A user just tried to transfer money but has insufficient funds.
The user is likely frustrated — they expected the transfer to go through. Our UX copy guidelines say: be empathetic, avoid blame, and always suggest a next step.
Give me 5 error message variations. Each should be under 80 characters and include a clear action the user can take. Use a friendly but professional tone — no emojis, no slang.
Here's an example of our current copy style: "Your payment is on its way. We'll notify you when it arrives."
The second prompt gives AI everything it needs: product context, user emotional state, constraints, format, and a voice example. The output will be dramatically better.
Five Prompt Templates for UX Designers
1. Research Synthesis Prompt
I conducted [number] user interviews for [product/feature].
Here are my raw notes:
[paste notes]
Analyze these notes and give me:
1. Top 5 recurring themes with supporting quotes
2. Pain points ranked by frequency
3. Unexpected insights that appeared in 2+ interviews
4. Suggested follow-up questions for the next round
Format each theme as: Theme Name — Description — Supporting quotes (with participant identifiers)
2. Persona Draft Prompt
Based on the following user research data, create a UX persona:
[paste research summary, survey data, or interview themes]
Include:
- Name, age range, job title, and a one-line bio
- Goals (3-4 bullet points)
- Frustrations (3-4 bullet points)
- Behaviors relevant to [product category]
- A "day in the life" scenario showing when they'd use our product
- A quote that captures their core attitude
Make this persona specific enough to guide design decisions, not a generic archetype.
3. UX Copy Generation Prompt
I need UX copy for [specific screen/component] in [product name].
Context: [describe the user's situation and what they're trying to do]
User emotion at this point: [e.g., excited, confused, frustrated, neutral]
Brand voice: [e.g., friendly and professional, playful, minimal and direct]
Design system constraints: [e.g., max character count, no jargon, action-oriented CTAs]
Give me [number] variations for each of these elements:
- [Heading/button/tooltip/error message/empty state — be specific]
For each variation, note why it might work better for different user segments.
4. Usability Test Script Prompt
I'm running a moderated usability test for [product/feature].
The main user flow being tested: [describe the flow step by step]
Participants: [describe target users]
Session length: [e.g., 45 minutes]
Create a usability test script that includes:
- A warm-up section (2-3 questions to make participants comfortable)
- Task scenarios (not instructions — describe situations, not steps)
- Follow-up probes for each task (what to ask when they hesitate or succeed)
- A wrap-up section with open-ended reflection questions
Important: tasks should describe realistic scenarios, not tell users what to click. Use "imagine you..." framing.
5. Design Critique Prompt
I'm working on [feature] for [product]. Here's a description of my current design:
[describe the layout, key components, user flow, and design decisions]
Act as a senior UX designer reviewing this work. Give me:
1. Three things that work well from a UX perspective
2. Three potential usability concerns with specific suggestions
3. Accessibility considerations I might have missed
4. One alternative approach I should consider and why
Be specific and reference design principles (Fitts's law, cognitive load, progressive disclosure, etc.) where relevant.
Common Prompting Mistakes UX Designers Make
Mistake 1: No user context. If you don't tell AI who the users are and what they need, it defaults to generic output. Always include user details.
Mistake 2: Asking for too much at once. "Design me a complete app" will get you shallow, generic results. Break complex tasks into focused prompts — research first, then personas, then individual screens.
Mistake 3: Not providing brand voice examples. AI can match any tone, but only if you show it what you want. Paste 2-3 examples of existing copy or describe the voice explicitly.
Mistake 4: Accepting the first output. AI output improves dramatically with follow-up. Say "make option 3 more concise" or "these feel too formal — make them warmer." Treat it as a conversation, not a single query.
Mistake 5: Not specifying constraints. Character limits, WCAG requirements, platform guidelines (iOS vs Android copy conventions) — tell AI about your constraints or it will ignore them.
Practice Exercise
Open ChatGPT or Claude and try this:
- Pick a product you're currently working on (or a well-known app like Spotify or Airbnb)
- Write a DUXE prompt asking AI to generate empty state copy for a screen where the user has no saved items
- Review the output, then send a follow-up prompt refining the best option
- Compare your DUXE prompt result with what you'd get from "write empty state copy"
The difference in quality will be obvious and immediate.
Key Takeaways
- Use the DUXE framework (Design context, User needs, eXpected output, Examples & constraints) for consistently better AI output
- Always provide user context, brand voice examples, and specific constraints in your prompts
- Break complex UX tasks into focused, sequential prompts rather than one giant request
- Treat AI as a conversation — refine outputs with follow-up prompts
- The five templates above cover research, personas, UX copy, usability testing, and design critique — your most common AI use cases