Check User Experience
Audit an interface against the Laws of UX and produce a prioritised report of real issues with concrete fixes. Report, don't fix, unless the user asks.
Workflow
1. Gather the surface
- Codebase: detect the stack, then find screens/routes, shared components (buttons, inputs, forms, nav, modals, lists, tables), design tokens, and loading/error states. Read the actual markup and styles, not just file names.
- Live URL / local build: open it in the browser, walk the main flows, take screenshots at desktop and mobile widths, measure hit targets and response times where possible.
- Screenshot: judge visually; mark code-only checks as "not verifiable".
- Identify the 3–5 primary user tasks (sign up, search, checkout, create X). Audit those flows hardest — Pareto applies to the audit too.
2. Check every law
Go through the table below. For each law, look for the listed signals. Only report concrete findings with a location; never pad the report with generic advice. If a law is satisfied, it goes in the "passing" list in one line.
3. Report (format below), then offer to fix the High items.
The laws: what to look for and how to fix
Interaction & effort
| Law | Flag when | Suggested fix |
|---|---|---|
| Fitts's Law — time to hit a target depends on size and distance | Tap targets < 44×44 pt (iOS) / 48×48 dp (Android) / 24×24 CSS px min (WCAG 2.5.8), ideally 44px; icon-only buttons with tiny padding; primary action far from where the eye/thumb is; destructive and primary buttons adjacent with < 8px gap | Increase padding or add ::after { inset:-8px } hit area; put primary CTA near the content it acts on, in thumb zone on mobile; separate destructive actions |
| Hick's Law — decision time grows with number/complexity of choices | Nav with > 7 top-level items; menus/selects with long unsorted lists; many equal-weight CTAs on one screen | Group, progressively disclose ("More"), add search/filter to long lists, sensible defaults, one primary CTA |
| Choice Overload | Pricing/plan pages with > 3–4 options and no recommendation; settings dumps; many filters shown at once | Highlight a recommended option, collapse advanced options, comparison table |
| Doherty Threshold — productivity soars when response < 400 ms | No loading/pending state on async actions; buttons don't disable or show progress on submit; full-page spinners for small loads; no optimistic updates; blocking network calls on input | Immediate feedback < 100 ms, skeletons, optimistic UI, delayed spinners (~300 ms), debounce search |
| Postel's Law — be liberal in what you accept | Strict input formats (phone, dates, card numbers rejecting spaces/dashes); case-sensitive emails; no trimming; validation errors that don't say how to fix; pasting blocked | Normalise input (trim, strip separators), accept multiple formats, inputmode/autocomplete attrs, helpful error messages |
| Tesler's Law — complexity can't be removed, only moved | User forced to do what the system could (re-enter known data, compute values, pick timezone/country that could be inferred) | Autofill, smart defaults, inference, remember last choices |
| Paradox of the Active User — users don't read manuals | Features only discoverable via docs/tooltips-on-hover; long onboarding tours before value; empty states with no guidance | Inline hints at point of use, actionable empty states, learn-by-doing onboarding |
| Flow | Unnecessary interruptions (modals, confirmations for reversible actions, forced re-auth); no keyboard shortcuts for power paths; lost focus after actions | Undo instead of confirm, preserve focus/scroll, keyboard support, minimise context switches |
Memory & cognition
| Law | Flag when | Suggested fix |
|---|---|---|
| Miller's Law / Chunking | Long unbroken forms; long strings shown raw (IDs, phone, card numbers unformatted); walls of text; nav with 10+ flat items | Section forms into groups/steps, format numbers in chunks, headings and short paragraphs |
| Cognitive Load | Dense screens with competing elements; jargon; inconsistent terminology for the same thing; decorative clutter | Remove non-essential elements, consistent labels, plain language |
| Working Memory | User must remember info from a previous screen (codes, earlier choices); placeholder-only labels that vanish while typing; comparisons across pages | Keep context visible (summaries, persistent labels, breadcrumbs), side-by-side compare |
| Mental Model | Controls that behave unlike their look (link styled as button doing navigation or vice versa); unusual metaphors; save model unclear (autosave vs manual) | Match the users' expectation; make system state explicit |
| Jakob's Law — users expect your site to work like others | Non-standard placement (logo not linking home, cart not top-right, search hidden); custom controls replacing native ones without parity; reinvented patterns | Follow platform conventions (HIG / Material / web norms); native elements first |
| Occam's Razor | Redundant steps, duplicate controls, features nobody's flow needs on the primary screen | Remove or move to secondary surfaces |
| Pareto Principle | Primary tasks buried while rarely-used features get prime space | Promote the 20% of features used 80% of the time |
| Parkinson's Law | Forms/tasks that take longer than needed; no autofill; no progress indicators that set expectations | Autocomplete, autofill attrs, show estimated time/steps |
| Selective Attention / Banner Blindness | Critical info styled like an ad or placed in banners/sidebars; important warnings blending in | Put key info inline in the task path, distinct but not ad-like |
| Cognitive Bias | Dark patterns: pre-checked opt-ins, confirmshaming, hidden costs, fake urgency, misleading defaults | Remove; honest defaults. Flag as High (ethics + legal risk) |
Perception (Gestalt)
| Law | Flag when | Suggested fix |
|---|---|---|
| Law of Proximity | Labels closer to the wrong input; equal spacing between unrelated and related items; helper text far from its field | Tighter spacing inside groups, larger between groups (spacing scale) |
| Law of Similarity | Same-looking elements behaving differently (clickable and static text identical); inconsistent button styles for same action level | Consistent visual language per role |
| Law of Common Region | Related controls not grouped; cards without boundaries mixing content | Cards, sections, background regions for groups |
| Law of Uniform Connectedness | Steps/related items not visually linked | Lines, connectors, shared containers for steppers/timelines |
| Law of Prägnanz | Overly complex icons/illustrations, ambiguous shapes | Simplify forms; recognisable icons with labels |
| Aesthetic-Usability Effect | Visual inconsistency (mixed radii, random colors, misaligned grid, inconsistent type scale) that erodes trust | Tokenise spacing/type/color; align to a grid |
Sequence, motivation & memory of experience
| Law | Flag when | Suggested fix |
|---|---|---|
| Von Restorff Effect | Primary CTA not visually distinct; everything highlighted so nothing stands out; reliance on color alone | One emphasised primary action per view; use shape/size/icon too (accessibility) |
| Serial Position Effect | Key nav items in the middle of long lists; most important info buried mid-page | Put key items first and last (e.g., Home first, Profile/Cart last in tab bar) |
| Peak-End Rule | Weak or missing success states (form submit ends with nothing or a plain redirect); harsh error dead-ends; abrupt flow endings | Clear, satisfying confirmation; helpful recovery on errors; polish the final step |
| Goal-Gradient Effect | Multi-step flows without progress indication; progress that starts at 0 | Stepper/progress bar, show steps remaining, give a head start |
| Zeigarnik Effect | Incomplete tasks forgotten (no drafts, lost form data on navigation, no profile-completion nudges) | Autosave drafts, "continue where you left off", completion checklists |
Severity
- High — blocks or seriously slows a primary task, causes errors/data loss, accessibility failure, or a dark pattern.
- Medium — noticeable friction or confusion on a common path.
- Low — polish; secondary screens.
Report format
# UX Audit — <app/screen>
Scope: <what was audited> · Method: code / live / screenshot
Score: <passing laws>/30 · High: n · Medium: n · Low: n
## Top 3 to fix first
1. ...
## Findings
### [High] Fitts's Law — Icon buttons too small
Where: src/components/Toolbar.tsx:42 (also Header.tsx:18)
Issue: 20×20px icon buttons with no padding; mis-taps on mobile.
Fix: add `p-3` (44px target) or an `::after` hit-area extension.
```(code snippet if short)```
(…one block per finding, grouped by severity, High first…)
## Passing
- Jakob's Law — standard nav and logo-home pattern.
- ...
## Not verifiable
- Doherty Threshold — needs a running app to measure response times.
Rules
- Every finding needs a law, a location (
file:line, URL + element, or screenshot region), the observed problem, and a specific fix. No location → don't report it. - Merge duplicates: one finding per root cause, listing all locations.
- Don't restyle or change the app's personality in fixes; suggest the smallest change that solves the issue.
- Respect platform conventions (iOS HIG, Material, web) when judging Jakob's Law and target sizes.
- Overlap with accessibility (contrast, focus, labels) is fine — note it under the relevant law, but don't turn this into a full WCAG audit.
- After the report, ask whether to fix the High (or all) findings. If motion/feedback issues are found (Doherty, Peak-End, Goal-Gradient), mention
/animate-this-appcan address them.