UX/UI accessibility audit
UX/UI Accessibility Audit
An evaluation of the AESOP AI evaluation platform against established usability and accessibility standards, for users with low technical literacy.
WCAG 2.2Nielsen NormanW3C COGAGOV.UKPlain LanguageWAI-AGEISO 9241Universal Design
01
Overall assessment
3.0
of 5.0
Functional, but not inclusive
AESOP delivers a capable evaluation platform with strong visual-design fundamentals. But its reliance on domain-specific jargon, military terminology, complex multi-path navigation, and an assumed technical literacy creates significant barriers for non-expert users. The biggest opportunity is language simplification and guided onboarding.
Needs improvement for low-literacy accessibility02
Findings by dimension
ANavigation Clarity
3.5 / 5
Strengths
- Persistent sidebar visible on all pages
- Back links on progress and completion pages
- Multi-step wizards with step indicators and Next / Back buttons
- Dashboard provides clear entry points via stat cards
Issues
- Nine sidebar items with no grouping or progressive disclosure COGA Obj.2
- Four overlapping evaluation paths (Cave, Investigate, Rubric, Evaluate) needing conceptual understanding to choose NNG #6
- No breadcrumbs on most pages WCAG 2.4.8
- No search for pages or features WCAG 2.4.5
- Welcome page assumes you already understand the three evaluation modes
- Inconsistent onboarding: separate Welcome and Dashboard flows
BLanguage Simplicity
2.0 / 5
Strengths
- Placeholder text gives concrete examples
- “How it Works” sections attempt explanation
- Some tooltips provide contextual definitions
Issues
- Heavy jargon with no definition: Red-Team, Black-Box, Adversarial, Resilience Score, Governance Documents WCAG 3.1.3
- Aggressive / military button labels: “Launch Red Team,” “Generate Attack Scenarios,” “Breached / Defended” Plain Language
- Technical API concepts exposed: POST requests, JSON payloads COGA Obj.3
- Abbreviations without expansion: AESOP, EMI, HIPAA, GDPR, FERPA WCAG 3.1.4
- Abstract capability names that don’t say what they do
- Reading level ~grade 12+ (target grade 6–8) WCAG 3.1.5
CError Prevention & Recovery
3.0 / 5
Strengths
- Form validation on submit with toast notifications
- File-upload validation (type, size limits)
- Disabled buttons and loading spinners prevent double-submission
- Required fields indicated
Issues
- Transient toasts disappear before slow readers finish COGA Obj.5
- No “check your answers” summary before starting GOV.UK
- No confirmation before launching costly evaluations WCAG 3.3.4
- Generic error messages with no corrective action WCAG 3.3.3
- No undo for started evaluations
- No indication of time or cost before starting
DCognitive Load
2.5 / 5
Strengths
- Progressive disclosure via collapsible sections
- Step-by-step wizard for the unified evaluation flow
- Report sections expandable / collapsible
- Dark mode reduces visual fatigue
Issues
- Five capability toggles with complex availability logic on Step 2 NNG #8
- Reports hold massive nested content (dimensions, scenarios, transcripts)
- Multiple evaluation types with subtle differences create choice paralysis COGA Obj.5
- About page has 5 tabs and 8 agent descriptions
- Score visualizations use 4 systems (% , ring, bar, tier)
- All features visible at once — no progressive onboarding
EVisual Hierarchy & Affordances
4.0 / 5
Strengths
- Clear hierarchy with consistent heading sizes
- Interactive elements visually distinct
- Consistent score color-coding
- Icons paired with text labels
- Consistent border, spacing, rounding system
- Clear status badges
Issues
- Some icon-only actions without text labels COGA Obj.1
- Help toggle is a small 40px button at the screen edge WAI-AGE
- Dark-mode muted purple ~3.8:1 contrast WCAG 1.4.3
- No documented keyboard focus indicators WCAG 2.4.7
FTrust & Transparency
3.0 / 5
Strengths
- Real-time progress indicators during evaluations
- Clear completion states with report access
- “How it Works” sections explain the process
- About page provides context and methodology
Issues
- No “before you begin” page (requirements, time, expected outcome) GOV.UK
- No cost indication before starting ISO 9241-110
- Help panel (RAG chat) hidden by default — most users won’t find it
- No “contact support” in the main flows COGA Obj.7
- Terms and Privacy buried in About tabs
03
Top ten recommendations
| # | Recommendation | Impact | Effort | Standards |
|---|---|---|---|---|
| 1 | Simplify terminology throughoutDefine every domain term on first use. Rename aggressive terms: “Red Team” → “Security Test,” “Attack Scenarios” → “Challenge Scenarios,” “Breached” → “Needs Improvement.” | High | Low | WCAG 3.1.3 · COGA 3 · Plain |
| 2 | Add “before you begin” start pagesTell users what materials they need, how long it takes, and what they’ll receive. The GOV.UK start-page pattern. | High | Low | GOV.UK · ISO 9241 |
| 3 | Unify the evaluation entry pointReplace four separate flows (Evaluate, Cave, Rubric, Investigate) with one guided wizard that asks “what do you want to learn?” then configures the right evaluation. | High | High | COGA 2 · NNG #6 · GOV.UK |
| 4 | Confirm before launch“You’re about to start a Quality Analysis. This typically takes 5–10 min. Here’s what you entered: [summary]. Start?” | High | Low | WCAG 3.3.4 · COGA 4 |
| 5 | Replace transient toasts with inline errorsPersistent, inline field-level errors with specific corrective instructions. Never rely solely on timed notifications. | High | Medium | WCAG 3.3.1 · COGA 4 · GOV.UK |
| 6 | Add contextual help on every fieldEvery form field gets a “?” icon explaining in plain language what the field means and why it matters. | Medium | Medium | COGA 7 · NNG #10 |
| 7 | Group sidebar navigationOrganize into sections — Evaluate (all eval types), Results (history, reports), Learn (guides, about) — cutting visible items from nine to 3–4 groups. | Medium | Medium | COGA 2 · NNG #8 |
| 8 | Add a visible “need help?” contact linkDisplay on every page, especially error states and progress pages; include email, chat, or documentation. | Medium | Low | COGA 7 · Universal |
| 9 | Fix dark-mode contrast ratiosMuted purple text (#a89ec0) on dark backgrounds falls below WCAG AA 4.5:1. Lighten to ~#c4b9d4. | Medium | Low | WCAG 1.4.3 · WAI-AGE |
| 10 | Add a guided onboarding tourFirst-time users walk through the key features — what AESOP is, how to start, where results live. Dismissable, re-accessible from help. | Low | High | ISO 9241 · COGA 1 |
04
Priority matrix — impact vs effort
Quick wins — do first
#1Simplify terminology
#2“Before you begin” pages
#4Confirmation before launch
Strategic — plan & invest
#3Unify evaluation entry
#5Inline error messages
Easy additions
#8“Need help?” link
#9Fix contrast ratios
Consider later
#6Field-level help tooltips
#7Group sidebar nav
#10Onboarding tour
05
Standards & frameworks applied
WCAG 2.2
Web Content Accessibility Guidelines — Understandable (3.1), Input Assistance (3.3), Navigable (2.4).
Web Content Accessibility Guidelines — Understandable (3.1), Input Assistance (3.3), Navigable (2.4).
—
3.1.3, 3.1.5, 3.3.1–4, 2.4.5–8, 1.4.3
Nielsen Norman
Ten usability heuristics plus NNG research on low-literacy users.
Ten usability heuristics plus NNG research on low-literacy users.
—
#1, #3, #5, #6, #8, #9, #10
W3C COGA
“Making Content Usable for People with Cognitive and Learning Disabilities.”
“Making Content Usable for People with Cognitive and Learning Disabilities.”
—
Objectives 1–7
GOV.UK principles
One thing per page, start pages, check answers, confirmation pages, reading age 9.
One thing per page, start pages, check answers, confirmation pages, reading age 9.
—
Start pages · error summary
Plain Language
U.S. Federal Plain Writing Act — active voice, short sentences, everyday words.
U.S. Federal Plain Writing Act — active voice, short sentences, everyday words.
—
Grade 6–8 reading level
WAI-AGE / ISO 9241
Guidance for older users; interaction principles — self-descriptiveness, controllability, error tolerance.
Guidance for older users; interaction principles — self-descriptiveness, controllability, error tolerance.
—
≥4.5:1 contrast · 44px targets
06
Plain-language rewrites
Red-Team
→
Security Test / Safety Check
Black-Box Discovery
→
Blind Test / Unknown System Test
Adversarial
→
Challenge / Stress Test
Resilience Score
→
Safety Score
Governance Documents
→
Policy Documents / Rules & Guidelines
Vital Context
→
Background Info / Additional Details
System Instructions
→
AI Instructions / Your AI’s Rules
Chatbot Endpoint
→
Chatbot URL / Where to reach your chatbot
Launch Red Team
→
Start Security Test
Generate Attack Scenarios
→
Create Test Challenges
Breached / Defended
→
Failed–Needs Improvement / Passed–Protected
Cave of Shadows
→
Security Testing (drop the allegory in labels)
EMI Certification
→
AESOP Quality Rating (expand the acronym)
Method. This audit was conducted by analyzing the full AESOP frontend — 35+ pages, navigation and form components, error handling, and user-facing text. Each dimension was scored against criteria from the eight standards above, reflecting the deployed app as of March 2026. It assesses the UX/UI experience for limited-literacy users only — not code quality, performance, or security. The pattern across all six dimensions is the same: sound structure undone by inaccessible language and unguided choice. The platform is more usable than its words allow.