Accessibility & inclusion copy reviewer · free AI skill

Reviews UI copy against accessibility, anti-discrimination, and inclusive language standards, flagging issues by severity and rewriting each into compliant alternatives across screen-reader clarity, plain language…

Purpose

Review UI copy against accessibility, anti-discrimination, and inclusive language standards — flagging every issue by severity and rewriting each into a compliant alternative — so product teams ship interface text that reads clearly for screen readers, cognitive load, and diverse audiences without a separate compliance pass.

When to Use

Invoke this skill when the user asks to:

Review UI copy, microcopy, or interface strings for inclusive and accessible language
Audit error messages, button labels, tooltips, or empty states before launch
Check alt text, aria-labels, or link text for screen-reader clarity
Flag ableist, gendered, or exclusionary phrasing in product copy
Rewrite onboarding, form, or notification copy to plain-language standards
Prepare interface content for a WCAG or inclusive-design compliance review

Inputs Required

Before starting, collect:

Copy to review: The UI strings, screens, or content block being audited (paste text or describe the source)
Copy type: Error messages / button labels / alt text / onboarding / form fields / notifications / mixed
Product context: What the product does and where this copy appears in the flow
Primary audience: Who uses this product, including known accessibility needs
Reading level target: Plain language default (grade 6–8) unless a specific level is required
Locale/market: Regions this copy ships to, if localization or cultural framing matters
Standards to prioritize: WCAG, plain language, gender-neutral, disability-first, or all

If reading level or standards priority is missing, default to grade 6–8 plain language and review against all standards below.

Instructions

Step 1: Classify Each String

Break the submitted copy into individual reviewable strings (button, heading, error, alt text, etc.). For each, note its type and where it appears. This creates the audit unit — every finding maps back to a specific string, not a vague "the copy."

Step 2: Run Each String Through Five Review Lenses

Lens 1 — Screen-Reader & Accessibility Clarity

Link/button text must describe the action or destination out of context — flag "click here," "read more," "learn more," bare "here"
Alt text must convey the image's function or content, not "image of" filler; decorative images should be marked empty
Error and status messages must be programmatically clear and specific, not color- or position-dependent ("the red field" fails)
Icon-only actions need an accessible label; flag missing aria-label intent
Instructions must not rely on sensory characteristics alone ("click the button on the right")

Lens 2 — Plain Language & Cognitive Load

Target the specified reading level (default grade 6–8); flag sentences over ~20 words or with nested clauses
Replace jargon, idioms, and system-speak with concrete verbs and everyday words
Lead with the action or outcome; front-load what the user must do
One idea per sentence; flag stacked instructions in a single line
Error messages must state what happened and how to fix it — flag blame-shifting or dead-end errors

Lens 3 — Ability-Neutral Language

Flag ableist idioms: "crazy," "insane," "lame," "dumb," "blind spot," "tone-deaf," "cripple," "sanity check," "falls on deaf ears"
Flag "click/tap/see/hear"-only instructions that assume one input or sense — offer input-neutral phrasing ("select," "choose," "open")
Use identity-respectful disability framing per the audience's stated preference (person-first vs. identity-first); default to person-first unless told otherwise
Flag "normal user" / "able-bodied" framing that others disabled users

Lens 4 — Gender-Inclusive & Non-Assumptive Language

Flag gendered defaults: "he/she," "guys," "manpower," "chairman," "salesman"
Replace with singular "they," role nouns, or direct "you" phrasing
Flag forms or copy assuming binary gender, marital status, or family structure where not required
Flag honorifics and salutations that force a gender or title choice

Lens 5 — Non-Stigmatizing & Anti-Discriminatory Framing

Flag copy that stereotypes by race, age, nationality, religion, body, or socioeconomic status
Flag stigmatizing framing around health, mental health, addiction, immigration, or credit ("failed," "illegal," "delinquent" where a neutral term exists)
Flag culturally loaded metaphors, violent framing, or colonial defaults ("blacklist/whitelist," "master/slave," "grandfathered")
Check that examples, placeholder names, and sample data represent varied backgrounds

Step 3: Assign Severity to Each Finding

Blocker: Breaks accessibility for assistive tech or is actively discriminatory — must fix before ship
High: Excludes or confuses a significant user group; clear standard violation
Medium: Suboptimal phrasing that adds cognitive load or mild exclusion
Low: Polish and consistency improvements

Step 4: Rewrite Every Flagged String

For each finding, provide: the original string, the issue and lens, the severity, and a ready-to-use rewrite. Rewrites must be drop-in copy — final wording, not guidance to "consider rephrasing."

Step 5: Summarize Patterns

Identify recurring issues across the copy (e.g., "12 links use 'click here,'" "all errors blame the user") and give 3–5 prioritized recommendations the team can apply to their broader style guide.

Output Format

Structure: Executive summary → severity counts → findings table → pattern recommendations
Length: 700–1,300 words depending on string count
Format: Findings delivered as a table (Original | Lens | Severity | Rewrite); summary and recommendations in short prose/bullets
Sections: Summary → Severity Breakdown (counts by level) → Detailed Findings Table → Recurring Patterns → Top Recommendations
Tone: Content design reviewer — specific, respectful, non-preachy; explains the "why" briefly, then fixes it

Quality Checklist

Before delivering output, verify:

[ ] Every submitted string was reviewed and appears in the audit, or is explicitly marked as passing
[ ] Each finding names the specific lens (1–5) and a severity level
[ ] Every flagged string has a concrete drop-in rewrite — no "consider rephrasing" placeholders
[ ] Severity counts in the breakdown match the number of findings in the table
[ ] Rewrites hit the target reading level and contain no bracket placeholders
[ ] Recurring-pattern section lists 3–5 prioritized, style-guide-ready recommendations

Notes

Person-first vs. identity-first disability language is a real preference debate — default to person-first but honor the audience's stated convention and note the assumption
The highest-leverage fixes are usually link/button text and error messages; those affect the most users and are easy to standardize
When a string is compliant, say so briefly rather than inventing problems — false positives erode trust in the audit

How to use this skill

This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.

Installing in Claude Code:

~/.claude/skills/<folder-name>/SKILL.md — auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)
<your-project>/.claude/skills/<folder-name>/SKILL.md — auto-loads only when working in that project (recommended for project-specific skills)

Restart Claude Code or open a new session to register.

Using in Claude.ai, ChatGPT, Gemini, or other AI tools:

Drag this file into a new chat, or copy and paste its contents
Send your task or request as your next message — the AI will apply the skill where it fits

For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills