Published on

25 ChatGPT Prompts for UX Designers (2026)

Authors
  • avatar
    Name
    PromptShelf Editorial
    Twitter

Most of a UX designer's week is not spent in Figma. It goes to research plans, interview scripts, synthesis notes, flow write-ups, microcopy, and the stakeholder updates that keep a project moving. ChatGPT will not design your screens or replace talking to real users. What it can do is take the writing and structuring work off your plate so you get more hours back for the thinking that actually needs you.

The 25 ChatGPT prompts for UX designers below are grouped into five stages of product work: user research, personas and journey mapping, wireframes and flows, UX writing, and usability testing plus stakeholder communication. Each one is copy-paste ready. Swap the bracketed details for your own project and you are set.

How to use these ChatGPT prompts for UX designers

Every prompt gives ChatGPT a role, a task, and an output format. That structure is what turns a vague answer into something you can use. The more specific your bracketed details, the sharper the result.

Two rules before you start. First, ChatGPT does not know your users, your product, or your analytics, so treat every draft as a first pass you edit, not a finished artifact. It will happily invent a research finding or a user quote you never gave it. Second, keep real user data out of prompts you send to a free tool you have not vetted. Strip names, emails, and anything sensitive from transcripts before you paste them. With that in mind, here are the prompts.

User research and discovery

Good research starts with sharp questions. These prompts help you plan studies and pull structure out of raw input.

1. Draft a research plan

Prompt: "You are a senior UX researcher. Write a lean research plan for a [mobile grocery app] feature study. Include the research question, objectives, method, participant criteria, sample size, and the key decisions this research should inform. Keep it to one page and use clear headings."

Use it to align your team before recruiting starts. Cut anything that does not tie back to a real decision.

2. Write interview questions

Prompt: "You are a UX researcher preparing a 30-minute user interview about [how people plan weekly meals]. Write 10 open-ended, non-leading questions ordered from warm-up to deeper probes. Avoid yes/no phrasing and avoid asking users to predict future behavior. Add a one-line note on what each question is trying to learn."

This gives you a script you can react to. Edit the wording so it sounds like you, not a survey.

3. Build a screener survey

Prompt: "You are a UX researcher recruiting for a usability study. Write a screener survey to find [small business owners who invoice clients at least twice a month]. Include qualifying and disqualifying questions, and hide the target answer so people cannot game it. Format as a numbered list with answer options."

A good screener saves you from wasting a session on the wrong participant. Always test it on a colleague first.

4. Turn interview notes into themes

Prompt: "You are a UX researcher doing thematic analysis. Group these raw interview notes into themes. For each theme, give a short name, a one-sentence description, and the note snippets that support it. Do not invent quotes or findings that are not in the notes. Notes: [paste your notes]."

This is the fastest way to move from a messy doc to a structure you can present. You still verify every theme against the source.

5. Draft a survey with balanced scales

Prompt: "You are a UX researcher. Write a 10-question post-launch survey measuring satisfaction and task success for [a new checkout flow]. Mix rating scales with two open-ended questions. Keep the language neutral and avoid double-barreled questions. Note which metric each question maps to."

Use it when you need signal at scale. Check that every scale is balanced and every question asks one thing.

Personas and journey mapping

Once you have research, these prompts help you shape and communicate what you learned.

6. Draft a persona from research

Prompt: "You are a UX designer building a persona from real research. Using these findings, draft a persona with goals, frustrations, context of use, and behaviors. Base every attribute on the findings and flag anything that is an assumption rather than evidence. Do not add demographic details I did not provide. Findings: [paste your synthesis]."

Personas drift into fiction fast. Forcing every line back to evidence keeps them honest.

7. Map a user journey

Prompt: "You are a UX designer. Build a journey map for [a first-time user setting up a budgeting app]. Break it into stages, and for each stage list the user's actions, thoughts, emotions, pain points, and opportunities. Format as a table with one row per stage."

This gives you a skeleton to fill with real quotes and data. It is a starting frame, not the finding itself.

8. Find friction in a current flow

Prompt: "You are a UX designer reviewing an existing flow. Here are the steps a user takes to [return an item in an online store]. For each step, list likely friction points and one hypothesis for why users might drop off. Mark which are worth testing versus fixing outright. Steps: [list the steps]."

Use it to build a hypothesis list before a usability test. Treat every point as a guess to validate.

9. Write a problem statement

Prompt: "You are a UX designer. Turn this research insight into a clear problem statement and three 'How might we' questions that open up solution space without prescribing a solution. Keep each one short and specific. Insight: [paste the insight]."

A tight problem statement keeps a kickoff from spiraling into feature debates. This gets you a first draft to sharpen.

10. Build an empathy map

Prompt: "You are a UX designer facilitating a synthesis workshop. Create an empathy map for [a nurse using a shift-scheduling tool] based on these observations. Fill the says, thinks, does, and feels quadrants using only what the observations support, and list gaps where we lack data. Observations: [paste them]."

Empathy maps are most useful when they show what you do not know yet. The gaps section is the point.

Wireframes, information architecture, and user flows

ChatGPT cannot draw, but it can help you think through structure before you open a canvas.

11. Outline a user flow

Prompt: "You are a UX designer. Write the step-by-step user flow for [booking a fitness class in a studio app], including the happy path and two edge cases (class full, payment declined). For each step, note the screen, the user's goal, and the primary action. Format as a numbered flow."

Use it to pressure-test a flow before wireframing. Add the edge cases your product actually hits.

12. Propose an information architecture

Prompt: "You are an information architect. Propose a site map for [a nonprofit's donation and volunteer website]. Group pages under clear top-level categories, keep navigation to a maximum of seven items, and explain your grouping logic in one line per category. Format as a nested list."

This gives you a structure to react to and test with card sorting. Do not ship it without validating the labels.

13. Write card-sort categories

Prompt: "You are a UX researcher planning a closed card sort. Given these [30 content items] for [a help center], propose 6 to 8 category labels users might sort them into, plus the items themselves as cards. Keep labels in plain user language, not internal jargon. Items: [paste the list]."

A closed card sort only works if your categories reflect how users think. This gets you a testable draft.

14. Draft acceptance criteria

Prompt: "You are a UX designer working with engineers. Write acceptance criteria for [a search filter component] in given-when-then format. Cover default state, applied filters, no results, and clearing filters. Keep each criterion testable and unambiguous."

Clear criteria cut the back-and-forth in build reviews. Read them against real edge cases before handoff.

15. Plan responsive breakpoints

Prompt: "You are a UX designer. For [a dashboard with a data table, filters, and a side nav], describe how the layout should adapt across mobile, tablet, and desktop. For each breakpoint, note what collapses, what stays, and the priority order of elements. Format as a table."

Use it to think through layout priority before you design three versions. The priority order is the useful part.

UX writing and microcopy

Words are part of the interface. These prompts speed up the copy that shapes how a product feels.

16. Write error message variants

Prompt: "You are a UX writer for a mobile banking app. A user tries to send a money transfer but has insufficient funds. Write three versions of the error message. Each must be under two sentences, avoid blame, avoid jargon and error codes, use plain language, and tell the user what they can do next. Format as a numbered list with a short label for the tone of each version."

Error states are where trust is won or lost. This gives you options to test instead of settling for the first line that works.

What this prompt actually returns (tested 2026-07-08)

We ran Prompt 16 on free ChatGPT with the example brief baked in (a mobile banking app where a user tries to send a transfer but has insufficient funds). Reproduced verbatim:

  1. Friendly: You don't have enough money in this account to send this transfer. Add money to your account or enter a smaller amount and try again.
  2. Calm: This transfer can't be completed because your available balance is too low. Deposit funds or reduce the transfer amount, then try again.
  3. Direct: There isn't enough money available to send this transfer. Add funds or send a smaller amount to continue.

Editor's notes. The model held every hard constraint: three variants, each under two sentences, no blame, no error codes or jargon, plain language, and each ends with a concrete next action. The tone labels are a useful touch, though "Friendly" and "Direct" land close enough that in practice you would test the "Calm" version against one of them rather than all three. What is missing is any product-specific path: a real banking app would link the user to a deposit or transfer screen, and none of these name where that action lives. Watch the assumption in "Add money to your account" too, since a user who is short on cash may not be able to, and that phrasing can read as tone-deaf. Treat these as a strong starting set, then wire each to the actual next step in your flow and test the wording with real users.

17. Draft empty-state copy

Prompt: "You are a UX writer. Write empty-state copy for [a task list in a project management app when a user has created no tasks yet]. Give a heading, one line of body copy, and a button label. Keep it encouraging without being cute, and make the next action obvious. Provide three options."

Empty states are a chance to orient a new user, not a dead end. Pick the version that points to the clearest first action.

18. Improve a confusing label

Prompt: "You are a UX writer. Users find the button labeled '[Sync now]' confusing in [a note-taking app]. Suggest five clearer alternatives, each with a one-line rationale for what it communicates. Keep labels to three words or fewer and match a calm, plain tone."

Small label changes fix outsized confusion. Test the top two in context before committing.

19. Write onboarding microcopy

Prompt: "You are a UX writer. Write copy for a three-step onboarding tooltip sequence for [a habit-tracking app]. Each step: a short heading and one sentence of body. Focus on what the user can do, not what the app is. Avoid exclamation marks and marketing language."

Onboarding copy should reduce effort, not sell. This gets you a draft to trim down further.

20. Rewrite for plain language

Prompt: "You are a UX writer focused on accessibility. Rewrite this interface text at roughly an 8th-grade reading level without losing meaning. Keep it concise, use active voice, and avoid idioms that do not translate. Return only the rewritten text. Text: [paste the copy]."

Plain language helps every user, not only those who need it most. Always check the rewrite kept the exact meaning.

Usability testing and stakeholder communication

The last stretch of a project is testing what you built and getting others to act on it.

21. Write usability test tasks

Prompt: "You are a UX researcher running a moderated usability test on [a food delivery app's reorder feature]. Write five realistic task scenarios framed as goals, not instructions, so you do not reveal the interface. Add the success criteria for each task. Format as a numbered list."

Task scenarios make or break a test. Framing them as goals keeps you from accidentally leading the participant.

22. Draft a test moderation script

Prompt: "You are a UX researcher. Write a moderation script for a 45-minute remote usability session, including the intro, consent reminder, think-aloud instructions, task transitions, and neutral probing phrases. Keep the probes non-leading. Format with clear section headings."

A script keeps your sessions consistent and your bias in check. Rehearse it once before the first participant.

23. Synthesize test findings

Prompt: "You are a UX researcher. Summarize these usability test observations into findings ranked by severity (critical, serious, minor). For each, give the issue, the evidence, and a recommendation. Do not include issues that are not supported by the observations. Observations: [paste them]."

This turns a wall of notes into a prioritized list a team can act on. You confirm each severity rating yourself.

24. Write an executive summary

Prompt: "You are a UX lead presenting to executives who are not designers. Summarize this research into a one-page brief: the question, what we learned, the three most important implications, and the recommended next step. Lead with the decision, not the method. Research: [paste your summary]."

Executives want the decision first. This gives you a draft framed for a reader who skims.

25. Handle design critique feedback

Prompt: "You are a UX designer preparing for a design review. Given this feedback from stakeholders, sort each comment into: usability issue, business constraint, or personal preference. For each, suggest a calm, evidence-based response I can give. Feedback: [paste the comments]."

Sorting feedback by type keeps a review from turning into a taste debate. Use it to walk in prepared, not defensive.

Tips for getting better results

Give ChatGPT constraints, not just topics. "Write onboarding copy" gets you filler. "Write three onboarding tooltips, each under 12 words, focused on the user's first action" gets you something you can almost ship.

Feed it your real inputs. Paste an actual (anonymized) interview transcript or a real set of notes instead of describing them. The output is only as grounded as what you give it.

And read everything before it reaches a user or a stakeholder. ChatGPT will invent a statistic, a quote, or a finding with total confidence. You are the designer. It is the assistant that types fast.

FAQ

Can ChatGPT do UX research for me?

No. It cannot recruit participants, run sessions, or observe real behavior, and it should never generate fake users or findings. What it does well is the scaffolding around research: draft plans, interview scripts, screeners, and synthesis structures that you fill with real data and verify yourself.

Is it safe to paste user research into ChatGPT?

Be careful. Strip names, emails, and any sensitive details from transcripts before pasting, and do not upload anything covered by an NDA or privacy agreement into a free tool you have not vetted. Anonymize first, and when in doubt, keep it out of the prompt.

Will using AI make my UX writing sound generic?

Only if you ship the first draft. The default output is safe and bland. Use ChatGPT to generate options and break a blank page, then edit in your product's voice and the specific detail that makes copy feel human. That editing pass is what keeps the work yours.

Which ChatGPT version do I need for these prompts?

The free version handles all 25 fine. Paid tiers give you higher limits and newer models, which help if you run many prompts a day or paste long transcripts, but none of these prompts require a paid plan to work.

How do I get ChatGPT to match my team's tone?

Show it examples. Paste a few pieces of microcopy or a style guide excerpt and ask it to mirror your voice, sentence length, and formality. The more of your real writing it sees, the closer the drafts land.

Start with three

You do not need all 25 today. Pick the three tasks that eat your week: maybe the interview script, the synthesis pass, and the error-message copy. Save those prompts, use them on your current project, and add more as they prove useful. The goal is not to design with AI. It is to spend less time on the writing around the work and more time on the design only you can do.

Related: more prompts by profession