AI can generate a polished paragraph in seconds. That does not automatically make it useful to a designer.
The difference usually comes down to the input. A vague request such as “Give me UX ideas for this website” tends to produce a vague list of familiar ideas. Give ChatGPT the product context, users, business goal, constraints, evidence, and desired output, and the conversation becomes much more useful.
I have found this particularly valuable in practical UI/UX work. ChatGPT is good at helping structure messy thinking, generate alternatives, challenge assumptions, turn research notes into working material, and review interfaces from different perspectives. It is much less reliable when asked to invent research findings or make design decisions without enough context.
OpenAI’s current guidance also recommends being clear and specific, supplying relevant context, defining the desired output, and refining prompts based on the response rather than expecting one perfect prompt to solve everything.
The ten prompts below are designed around that workflow. Copy them, replace the bracketed sections, and then refine the result with your actual project information.
How designers should use ChatGPT in a UX workflow
Before getting to the prompts, there is one important shift in mindset.
Do not treat ChatGPT as a replacement for your design process. Treat it as a fast thinking partner that can help you inspect, organize, question, and develop your work.
Give the prompt enough project context
Compare these two requests:
“Give me ideas for a dashboard.”
and:
“I am designing a B2B inventory dashboard for warehouse managers who check stock levels several times a day. The primary task is identifying low-stock products quickly. The interface is desktop-first and needs to support 500+ SKUs. Suggest three information architectures…”
The second prompt gives the model something to reason about.
OpenAI recommends providing the necessary context and being specific about the desired outcome, format, tone, and constraints.
For UX work, useful context can include:
- Product type
- Target users
- Primary task
- Business goal
- Platform
- Existing problems
- Research evidence
- Technical limitations
- Content constraints
- Accessibility requirements
- What you have already tried
The more important the design decision, the more important the context.
Ask for analysis before asking for solutions
This is one of the biggest improvements you can make to AI-assisted design work.
Instead of immediately asking:
“How should I redesign this checkout?”
ask:
“Review this checkout flow. First identify usability problems, their likely causes, and the evidence needed to validate each problem. Do not propose solutions yet.”
That separation matters.
If you ask for problems and solutions in one sentence, AI can jump straight to familiar interface patterns. You may get a carousel, progress indicator, larger button, simplified form, or modal because those are common design solutions, not because the evidence supports them.
A better workflow is:
Observe → Analyze → Question → Generate → Evaluate
ChatGPT is particularly useful in the middle of that process.
Treat the first answer as a working draft
OpenAI explicitly recommends iterative refinement rather than searching for one perfect prompt.
For designers, this means your second prompt can be more important than your first.
After receiving an answer, try:
- “Which assumptions are you making?”
- “Separate evidence from interpretation.”
- “Give me three alternatives.”
- “What would make this recommendation wrong?”
- “Challenge your previous answer.”
- “Now prioritize these by user impact.”
- “Turn this into a testable hypothesis.”
That is much closer to how an experienced designer works with another designer.
Prompt 1: Turn messy research notes into UX insights
User research rarely arrives in a perfectly organized document.
You may have interview notes, screenshots, observations, quotes, survey responses, and random comments collected during meetings.
ChatGPT can help structure that material, but the prompt needs a strong evidence boundary.
When to use it
Use this when you have raw interview notes or research observations and need to identify patterns without pretending that every comment is a validated insight.
Copy and paste prompt
Act as a senior UX researcher helping me analyze qualitative research.
I will provide raw research notes below.
Your job is to organize the evidence without inventing information.
Context:
- Product: [PRODUCT]
- Target users: [USERS]
- Research goal: [GOAL]
- Number of participants: [NUMBER]
- Research method: [INTERVIEWS / USABILITY TESTS / SURVEY / OTHER]
Research notes:
"""
[PASTE NOTES HERE]
"""
Analyze the notes in this order:
1. List the strongest recurring observations.
2. Group similar observations into themes.
3. For each theme, provide:
- What participants actually said or did
- The pattern you see
- A possible UX implication
- What evidence is still missing
4. Separate direct evidence from your interpretation.
5. Identify contradictions between participants.
6. Identify findings that appear only once and should not be treated as a general pattern.
7. Suggest 5 follow-up research questions.
Do not invent quotes, user motivations, demographics, or findings that are not supported by the notes.
Use a table where useful.The last instruction is important.
AI can make a neat research summary look more certain than the underlying evidence actually is. Keeping observations, interpretation, and missing evidence separate makes the output much safer to use.
Prompt 2: Generate user interview questions
A common research mistake is writing questions that quietly contain the answer.
For example:
“Would you find it easier if we added filters?”
That is already suggesting a solution.
A better research question investigates the problem before introducing the proposed interface.
When to use it
Use this before stakeholder interviews, user interviews, discovery sessions, or usability research.
Copy and paste prompt
Act as a senior UX researcher.
I am planning user interviews for the following product:
Product:
[DESCRIBE PRODUCT]
Target users:
[DESCRIBE USERS]
Research objective:
[WHAT I NEED TO LEARN]
Current assumptions:
[LIST ASSUMPTIONS]
Known product problems:
[LIST KNOWN PROBLEMS]
Create a user interview guide that helps me investigate the problem without leading participants toward a particular solution.
Structure the guide as:
1. Warm-up questions
2. Questions about current behaviour
3. Questions about goals and motivations
4. Questions about frustrations and workarounds
5. Questions about decision-making
6. Questions about the existing product
7. Questions that test my assumptions
8. Closing questions
For every question:
- Keep the wording conversational.
- Avoid leading questions.
- Avoid suggesting a solution.
- Explain what the question is intended to learn.
Then identify 5 questions I should NOT ask because they could introduce bias, and explain why.This works because the prompt asks ChatGPT to design the research around learning rather than confirmation.
You can then take the generated questions and edit them based on the actual context of your participants.
Prompt 3: Audit a website for UX problems
A UX audit becomes more useful when the reviewer knows what to inspect.
Instead of asking ChatGPT whether a website is “good,” define the dimensions of the review.
When to use it
Use this for an early website review, competitor analysis, redesign discovery, or a pre-launch audit.
If you can provide screenshots, page content, URLs, analytics information, or usability observations, the review becomes more grounded.
Copy and paste prompt
Act as a senior UX consultant conducting a website UX audit.
Website/product:
[DESCRIBE WEBSITE]
Target audience:
[DESCRIBE USERS]
Primary business goal:
[GOAL]
Primary user goals:
[USER GOALS]
Important pages:
[LIST PAGES]
Known problems:
[LIST IF ANY]
Review the information I provide and audit the experience across:
1. Clarity of value proposition
2. Information architecture
3. Navigation
4. Content hierarchy
5. Calls to action
6. Forms
7. Mobile usability
8. Accessibility
9. Trust and credibility
10. User journey continuity
11. Error prevention and recovery
12. Conversion friction
For each issue provide:
- Issue
- Evidence
- Why it may hurt the user
- Severity: Critical / High / Medium / Low
- Confidence: High / Medium / Low
- Recommended next step
- How I could validate the issue
Do not invent user behaviour or analytics data.
If the evidence is insufficient, explicitly say what information is missing.
At the end, give me the 5 issues I should investigate first, based on user impact rather than visual preference.That final sentence is useful because it stops the review from becoming a list of cosmetic observations.
A UX audit should not become a colour-and-spacing critique simply because those are easy things to notice.
Prompt 4: Improve website information architecture
Information architecture is one area where designers can use AI effectively for exploration.
The goal is not to let AI decide the sitemap. The goal is to generate alternatives and expose assumptions.
When to use it
Use this when a website has too many pages, confusing navigation, overlapping services, or unclear content categories.
Copy and paste prompt
Act as a senior information architect.
I am restructuring the information architecture for:
Website:
[WEBSITE TYPE]
Business:
[DESCRIBE BUSINESS]
Target users:
[USERS]
Current navigation:
[PASTE CURRENT NAVIGATION]
Current pages:
[LIST PAGES]
User goals:
[LIST GOALS]
Business priorities:
[LIST PRIORITIES]
Known problems with the current structure:
[PROBLEMS]
Create three alternative information architecture models.
For each model provide:
1. Primary navigation
2. Secondary navigation where needed
3. Page hierarchy
4. Which pages should be merged
5. Which pages should be renamed
6. Which pages may be unnecessary
7. Potential user confusion
8. Strengths and weaknesses
Use familiar language rather than internal company terminology.
Do not select a final architecture for me.
After presenting the three models, identify the assumptions that I should validate with users before choosing one.This is a much better use of AI than asking:
“Create the perfect sitemap.”
There is rarely a single perfect sitemap without understanding the audience, content, business structure, and user tasks.
Prompt 5: Create UX personas from real evidence
Personas can become fictional characters surprisingly quickly.
A good prompt should prevent ChatGPT from filling missing information with plausible-sounding demographics.
When to use it
Use this when you have research evidence and need to organize user groups around behaviour, goals, needs, and frustrations.
Copy and paste prompt
Act as a UX researcher helping me create evidence-based personas.
I will provide research material below.
Product:
[PRODUCT]
Research participants:
[NUMBER AND TYPE]
Research notes:
"""
[PASTE RESEARCH]
"""
Create 3 to 5 provisional personas only if the evidence supports meaningful differences between user groups.
For each persona include:
- Persona name
- Behavioural profile
- Primary goals
- Main tasks
- Frustrations
- Needs
- Common behaviours
- Context of use
- Product expectations
- Relevant research evidence
Do NOT invent:
- Age
- Income
- Job title
- Family status
- Location
- Technology preferences
- Personality traits
unless the research explicitly supports them.
After creating the personas, explain:
1. Which details are directly supported by evidence.
2. Which details are interpretations.
3. Where the research is insufficient.
4. Whether any personas should be merged.
5. What additional research would improve them.
Treat these as provisional research tools, not facts about all users.This is a useful distinction because a persona should help a design team make decisions. It should not become a fictional biography that nobody can verify.
Prompt 6: Review a UI design like a senior designer
This is one of the most useful prompts when you have a screenshot or design file you want to critique.
The trick is to request a structured critique instead of asking whether the interface “looks good.”
When to use it
Use it when you have a UI screen, prototype screenshot, or design concept and want another perspective before presenting it.
Copy and paste prompt
Act as a senior product designer reviewing a UI design before stakeholder presentation.
I will provide an image/screenshot of the interface.
Product context:
[PRODUCT]
Target user:
[USER]
Primary task:
[PRIMARY TASK]
Platform:
[WEB / MOBILE / TABLET]
Design goal:
[GOAL]
Review the interface across these areas:
1. Visual hierarchy
2. Layout and spacing
3. Typography
4. Interaction clarity
5. Information density
6. Navigation
7. Content hierarchy
8. Calls to action
9. Accessibility
10. Responsive considerations
11. Consistency
12. Potential usability problems
For each issue provide:
- What you observe
- Why it matters
- Severity
- Suggested improvement
- What evidence would validate the change
Separate objective observations from subjective design preferences.
Do not praise the design just to be positive.
Do not recommend changes simply because they are fashionable.
Focus on whether the interface helps the target user complete the stated task.This last instruction changes the quality of the critique.
A design can be visually unconventional and still work. A familiar design pattern can also fail if it does not fit the user’s task.
Prompt 7: Generate better UI copy and microcopy
Designers frequently receive placeholder copy from clients and then discover that the actual words break the layout.
AI can help generate variations quickly, especially for buttons, empty states, labels, errors, and onboarding messages.
When to use it
Use this when you have a UI element but the wording feels vague, too long, too technical, or inconsistent.
Copy and paste prompt
Act as a UX writer working with a product designer.
Product:
[PRODUCT]
Audience:
[AUDIENCE]
UI element:
[BUTTON / FORM / ERROR / EMPTY STATE / ONBOARDING / OTHER]
Current copy:
"[PASTE COPY]"
User action:
[WHAT THE USER IS TRYING TO DO]
Desired outcome:
[WHAT HAPPENS AFTER THE ACTION]
Brand voice:
[VOICE]
Constraints:
- Maximum [X] characters where applicable
- Plain language
- No unnecessary jargon
- Do not exaggerate claims
- Make the action clear
- Keep terminology consistent
Generate 8 alternatives.
For each option:
- Give the copy
- Explain the UX reasoning in one sentence
- Identify whether it is best for clarity, brevity, reassurance, or action
Then recommend the 2 strongest options based on the stated user task.
Do not invent product capabilities that I have not provided.The character limit is particularly useful for UI design.
A sentence that looks excellent in a document may become three lines inside a 320-pixel mobile button.
Give the model the constraint before it writes.
Prompt 8: Create usability test tasks
A usability test can fail before the participant arrives if the task is poorly written.
“Explore the website and tell us what you think” is usually too broad.
A good task gives participants enough context to act without telling them exactly where to click.
When to use it
Use this when planning moderated or unmoderated usability testing.
Copy and paste prompt
Act as a senior UX researcher designing a usability test.
Product:
[PRODUCT]
Target participant:
[USER TYPE]
Prototype or website:
[DESCRIBE EXPERIENCE]
Research objectives:
[WHAT I NEED TO LEARN]
Key flows:
[LIST FLOWS]
Create 5 to 8 realistic usability test tasks.
For each task provide:
- Scenario
- Participant task
- What success looks like
- What I should observe
- Possible usability signals
- Follow-up question
Rules:
- Do not tell the participant which button to click.
- Do not use interface labels unless the participant would naturally know them.
- Keep scenarios realistic.
- Do not turn tasks into instructions.
- Avoid leading participants toward a specific solution.
At the end, identify which tasks are most important if I only have 20 minutes with each participant.This is a good example of why context matters.
The same interface can require completely different research tasks depending on whether the participant is buying something, managing inventory, booking an appointment, or applying for a service.
Prompt 9: Turn a UX problem into design hypotheses
One of the most useful things AI can do for a designer is help convert vague observations into testable statements.
When to use it
Use this when someone says:
“Users seem confused.”
That sentence describes a problem but does not give the team something useful to test.
Copy and paste prompt
Act as a senior UX strategist.
I will give you a UX problem statement.
Product:
[PRODUCT]
Target user:
[USER]
Observed problem:
[PROBLEM]
Evidence:
[RESEARCH / ANALYTICS / OBSERVATION]
Business goal:
[GOAL]
Turn the problem into 5 testable UX hypotheses.
For each hypothesis provide:
- Hypothesis statement
- User behaviour we expect to change
- Proposed design direction
- Assumption behind the hypothesis
- What could prove it wrong
- Suggested usability test
- Suggested success metric
Do not present the proposed design as a proven solution.
Clearly distinguish:
1. What we know
2. What we suspect
3. What we could testThis structure is useful in design reviews because it changes the conversation from:
“I think we should move this button.”
to:
“We suspect users are missing this action because it is visually competing with another action. We could test whether stronger hierarchy improves task completion.”
That is a much healthier design discussion.
Prompt 10: Prepare a UX case study from project notes
A strong portfolio case study needs more than polished screens.
It needs a clear story: what happened, what you learned, what you changed, and why.
ChatGPT can help organize the material, but you should provide the actual project evidence.
When to use it
Use this after completing a project when you have research notes, design decisions, iterations, screenshots, and outcomes.
Copy and paste prompt
Act as a senior UX portfolio editor.
I am preparing a UX case study from a completed project.
Project:
[PROJECT NAME]
Product:
[PRODUCT]
My role:
[ROLE]
Team:
[TEAM]
Timeline:
[TIMELINE]
Business problem:
[PROBLEM]
User problem:
[USER PROBLEM]
Research:
[RESEARCH]
Key findings:
[FINDINGS]
Design iterations:
[ITERATIONS]
Final solution:
[SOLUTION]
Testing:
[TESTING]
Results:
[RESULTS]
Constraints:
[CONSTRAINTS]
Write a case study structure using:
1. Project overview
2. Problem
3. Context
4. My role
5. Research
6. Key findings
7. Design process
8. Important design decisions
9. Testing and iteration
10. Final solution
11. Outcome
12. What I learned
Rules:
- Do not invent metrics or research findings.
- Do not claim causation unless the evidence supports it.
- Separate my contribution from the team's contribution.
- Highlight decisions and reasoning rather than describing every screen.
- Point out where evidence is missing.
- Use clear, conversational professional language.
- Mark places where screenshots, research evidence, or metrics would strengthen the case study.
After the outline, give me 5 questions I should answer before publishing the case study.This is particularly useful for portfolio work because the AI can help organize the story while you remain responsible for the factual record.
How to get better results from these prompts
The ten prompts above are deliberately detailed, but you should not copy them forever without modification.
Your own project context is what makes them valuable.
Add constraints that actually matter
A good prompt might include:
- Platform: Mobile app
- Viewport: 390 × 844
- Audience: First-time users aged 18 to 30
- Task: Complete account registration
- Constraint: Must work with one hand
- Accessibility: WCAG 2.2 AA target
- Brand: Professional but approachable
- Technical: Existing React component library
Those details narrow the design space.
Without them, the model has to make assumptions.
Ask for alternatives when the decision is genuinely open
If you already know exactly what you want, ask for a focused critique.
If you are exploring a problem, ask for alternatives.
For example:
Give me three different approaches.
For each approach:
- Explain the interaction model.
- State the main advantage.
- State the main risk.
- Identify the user assumption.
- Tell me what I should test.This produces more useful design material than:
“Give me the best UX solution.”
There may not be one objectively best solution.
Use a second prompt to challenge the first answer
This is where AI becomes more useful than a single prompt pasted into a new chat.
After receiving a recommendation, try:
Now challenge your previous recommendation.
List:
1. The assumptions you made.
2. Situations where the recommendation would fail.
3. Alternative interpretations of the evidence.
4. What information you need before I make a decision.
5. The cheapest research method that could reduce the uncertainty.OpenAI’s current guidance recommends iterative refinement and treating prompting as a process rather than searching for one perfect instruction.
FAQs
Can ChatGPT replace a UX designer?
No. ChatGPT can assist with research synthesis, ideation, writing, critique, documentation, and structured analysis, but it does not replace the designer’s responsibility for understanding users, evaluating evidence, making trade-offs, testing designs, and working within real product constraints.
Its usefulness increases when the designer supplies strong context and evaluates the output critically.
What should I include in a ChatGPT prompt for UX design?
Include the product context, target users, task, business goal, constraints, evidence, desired outcome, and output format.
For example:
Product + users + problem + constraints + task + desired output
That is usually much more useful than simply telling ChatGPT to “act as a UX designer.”
Can I paste Figma screenshots into ChatGPT?
When image input is available in your ChatGPT experience, a screenshot can provide useful visual context for critique and analysis.
However, a screenshot does not provide everything a UX designer needs. It may not reveal interaction behaviour, analytics, research findings, responsive states, accessibility behaviour, or technical constraints.
Give the screenshot context such as the user, task, platform, and design objective.
Should I ask ChatGPT to design the entire app?
You can ask it to generate ideas or explore interface directions, but breaking a complex task into smaller stages generally produces more useful results.
A practical sequence is:
Research → problem framing → information architecture → flows → interaction ideas → UI copy → visual critique → testing plan
This also makes it easier to identify where the AI is making assumptions.
How do I stop ChatGPT from giving generic UX advice?
Give it specific context and ask it to show its assumptions.
Instead of:
“How can I improve this dashboard?”
try:
“This dashboard is used by warehouse managers checking stock levels during working hours. The primary task is identifying products below reorder thresholds. Review the attached screen and identify five usability risks. For each, explain the evidence, assumption, potential impact, and how I could test it.”
Specific context reduces the space for generic answers.
Is prompt engineering necessary for designers?
You do not need to learn a complicated technical system.
For most designers, good prompting means being clear about the task, giving relevant context, stating constraints, specifying the output format, and refining the response when it misses the mark.
OpenAI’s current guidance similarly emphasizes clear instructions, context, output requirements, and iterative refinement.
Conclusion
ChatGPT becomes much more useful to a UI/UX designer when the prompt describes the actual design problem instead of asking for generic “UX ideas.”
The ten prompts in this article cover research synthesis, interview planning, UX audits, information architecture, personas, UI critique, microcopy, usability testing, design hypotheses, and portfolio case studies. They are starting points, not rigid formulas.
The approach works best when you have real project context to provide. Research notes, user goals, screenshots, constraints, existing content, analytics, and technical limitations all give the model something concrete to work with. It works poorly when you ask AI to invent evidence or make decisions without knowing the users.
The most useful habit is simple: give ChatGPT better context, ask it to expose assumptions, then challenge the answer.
That turns prompting from a search for magic wording into a practical part of the design process.
