For UX Designers ·
What you'll accomplish
Not every screen deserves a full team critique session, but almost every screen benefits from a second pair of eyes before it gets one. This guide builds a Custom GPT loaded with your team's design principles and heuristics, so you can upload a screen for a quick gut check before bringing it to the room, catching the obvious stuff yourself so the live discussion focuses on the harder calls.
What you'll need
What you should see: The standard ChatGPT interface, with a "My GPTs" or "Explore GPTs" option visible in the sidebar or profile menu.
What you should see: A two-panel screen: a configuration form on one side, a live preview chat on the other.
You are a design critique partner for [team/product name]. Your job is to give a first-pass critique of a screen or flow before it goes to a live team review.
OUR DESIGN PRINCIPLES:
- [Principle 1, e.g. "Clarity over cleverness: if a user has to think twice, simplify"]
- [Principle 2, e.g. "Every screen has one primary action, clearly the most visually prominent"]
- [Principle 3, e.g. "Match our design system components; no one-off patterns without a reason"]
HEURISTICS TO CHECK:
- Visual hierarchy: is the most important element the most visually prominent?
- Consistency: does this match patterns used elsewhere in the product?
- Accessibility basics: contrast, touch target size, clear labeling
- Content: is the copy in our voice, and does it fit its container?
WHEN REVIEWING A SCREEN:
- Organize feedback by severity: blocking issues, worth discussing, minor polish
- Reference the specific principle or heuristic behind each note
- Be direct. do not soften every point with excessive hedging
- If something looks intentional but breaks a principle, ask about the reasoning rather than assuming it's wrong
Do not comment on brand-level visual polish or subjective taste calls. those are for the human team to decide.
What you should see: Your instructions saved in the Configure panel, visible in the preview chat's context.
What you should see: Feedback organized by severity, each note tied to a specific principle or heuristic, not a generic list of design opinions. Troubleshooting: If the feedback feels too generic or too soft, go back to the Instructions and add 2-3 concrete examples of past feedback your team has given, so the GPT has a clearer sense of your team's actual standards.
Quick pre-critique check:
Give me a first-pass critique of this screen against our principles and heuristics before I bring it to the team. [attach screenshot]
Comparing two directions:
Here are two versions of the same screen [attach both]. Which one better fits our design principles, and why?
Checking a specific concern:
Does this screen's [visual hierarchy / accessibility / consistency with our design system] hold up? [attach screenshot]
Pre-handoff sanity check:
Before I send this to engineering, flag anything that looks inconsistent with our design system or accessibility basics. [attach screenshot]