March 21, 2026

Design and UX Prompts That Make Feedback Useful

Design review is one of those things that sounds productive but often isn't. You share a screen, people say "maybe make it bigger" or "it feels a bit off," and you leave with a list of vague notes that could mean anything. Then the cycle repeats next sprint.

AI doesn't replace the judgment of a great designer. But it's remarkably good at giving you structured, specific, actionable feedback — fast. The catch is that you have to ask it the right way. A lazy prompt gets lazy feedback. A prompt that describes your context, your users, and your constraints gets something you can actually act on.

Here are the prompts I keep coming back to, organized by what you're trying to do.

UI Review

This is the most common use case and also the one where context matters most. Don't just paste a screenshot description and ask "is this good?" Tell the AI who the user is, what platform it's for, and what the screen is supposed to accomplish.

"Review this UI screen for a [mobile / web] [app type] targeting [audience]. The goal of this screen is [user goal]. Identify usability issues, visual hierarchy problems, and anything that might cause confusion. Be specific about what to fix and why."

"I'm designing a [checkout / onboarding / dashboard] flow. Here's the current layout: [description]. What are the three biggest friction points a first-time user would hit, and how would you fix each one?"

The second prompt forces prioritization, which is more useful than an exhaustive list of everything wrong.

UX Research

AI won't replace user interviews, but it's excellent for preparing for them, analyzing what you've already gathered, and synthesizing findings into something shareable.

"I'm planning usability tests for [feature / product]. The main questions I need to answer are [questions]. Write a 45-minute moderated interview script with tasks, follow-up probes, and a warm-up section. Avoid leading questions."

"Here are raw notes from 8 user interviews about [topic]: [paste notes]. Identify recurring themes, surface any surprising findings, and summarize the top 3 insights in plain language I could share with a non-design stakeholder."

That second one is particularly good. Getting patterns out of messy interview notes used to take a couple of hours. Now it takes minutes, and you still do the judgment call on what matters.

Design Systems

Design systems work is unglamorous and often neglected. These prompts help you think through decisions that would otherwise sit in someone's head.

"I'm building a component library for a [B2B SaaS / consumer app / internal tool]. We use [Figma / Storybook / etc.]. Write guidelines for how our Button component should handle states (default, hover, active, disabled, loading), including accessibility requirements and when to use primary vs. secondary variants."

"Our design system has grown inconsistently across three product teams. We have [X] spacing values, [Y] type styles, and duplicated components. Help me draft an audit checklist and a proposal for consolidating to a single token set without breaking existing products."

Wireframing

Before you open Figma, AI can help you pressure-test your structure. Describe what you're trying to build and let it poke holes before you've committed to a direction.

"I'm wireframing a [feature name] for [product type]. The user needs to accomplish [task] in as few steps as possible. Here's my current flow: [describe steps]. What's wrong with this flow? What would a simpler version look like?"

"Suggest three different layout approaches for a [page type, e.g. settings page / search results / user profile]. For each one, describe the structure, the tradeoffs, and what type of product or user it suits best."

Getting three options with explicit tradeoffs is way more useful than asking "what's the best layout?" — there is no universal best, and forcing the AI to acknowledge that keeps it honest.

Copy and Microcopy

Interface copy is one of the highest-leverage things you can work on and one of the most neglected. Good microcopy reduces support tickets, increases conversion, and makes your product feel trustworthy. Bad microcopy makes users feel stupid.

"Rewrite this error message to be clear, non-technical, and tell the user what to do next: [paste current message]. The user is [audience description] and probably hit this error because [likely cause]."

"Write 5 variations of the empty state for a [feature, e.g. notifications panel / saved items list]. The tone is [casual / professional / friendly]. Each version should explain what the space is for and nudge the user toward the action that fills it."

Always give the AI your brand tone and your audience. "Write button copy" gives you generic output. "Write button copy for a fintech app targeting first-time investors who are nervous about getting things wrong" gives you something you might actually ship.

Presentation and Stakeholder Handoff

Designing something is one job. Getting alignment on it is a completely different one. These prompts help you structure the story before you walk into the room.

"I'm presenting a redesign of [feature] to [audience: engineers / leadership / client]. The main change is [change]. Help me structure a 10-minute presentation that covers the problem we're solving, the decision we made, what we considered and ruled out, and what we need from this audience."

"Write a design handoff document for [component or feature]. Include: the intent behind each design decision, interaction states to implement, edge cases to watch out for, and open questions that engineering should flag back to design."

That last one is especially good for async teams. Writing it forces you to articulate decisions you made intuitively, and it gives engineers something to reference instead of pinging you for every ambiguity.

The tip that makes all of these better

Context is everything. Before you paste any of these prompts, ask yourself three things:

• Platform and constraints — mobile or web? Internal tool or consumer-facing? Brand guidelines?
• Audience — who are the actual users, and what do they know or not know?
• What "good" looks like — what are you optimizing for: speed, trust, delight, accessibility?

The more specific you are upfront, the less you'll have to argue with the output. AI feedback calibrated to your actual situation is genuinely useful. Generic AI feedback is noise.

And if you want to compare how Claude, ChatGPT, and Gemini each interpret the same design prompt — they often come at it from different angles, which is itself useful — that's exactly what Prompt Router is for.