A tool for checking language mismatches in web accessibility.
GitHub Repo • Usage • Contribute
Kizuki is a prototype accessibility auditing tool built on top of Google Lighthouse.
It adds custom checks to evaluate language mismatches between visible content and accessibility metadata.
This tool is designed to support analysis in multilingual environments, especially in non-Latin-script websites.
Kizuki analyzes:
- What language the visible content is in
- What language is used in:
altattributesaria-labelfields- form
labelelements
- Whether these languages match
- If accessibility text is missing, empty, or generic
The main entry point is:
node run-lighthouse.jsIt accepts either a CSV file containing multiple URLs.
Kizuki will:
- Run Lighthouse with custom audits
- Save results in the reports/ folder
- HTML (for visual inspection)
- JSON (for programmatic use)
You can find some sample reports already inside reports/.
For a list of URLs (CSV file):
node run-lighthouse.js --csv=websites.csv
CSV must contain one URL per line.
📁 Output
Kizuki stores all results in reports/. Each site produces two files:
example.com.json
example.com.html
The JSON can be parsed for metrics, and the HTML is Lighthouse's full visual report with added audits.
To define and register a new custom audit:
Create your audit in the audits/ folder.
Use existing files as a reference (e.g., alt-text-language-audit.js).
Edit index.js
Add your audit to the list of audits being exported.
Update custom-config.js
Register the new audit under the audits and categories fields.
Verify in run-lighthouse.js
Ensure the run uses custom-config.js (already set by default).
Your audit will now run for every URL, and results will appear in both .json and .html output.
This is an early-stage tool. Expect:
- limited language detection for mixed content
- brittle performance with dynamic or SPA websites
- minimal config options
It is intended as a research prototype, not a production-ready audit framework.
Ways to help:
- Add new language-sensitive accessibility checks
- Improve accuracy of language detection
- Support additional output formats
- Extend support for international scripts
Open a pull request or issue to get started.
Apache 2.0
Email: [email protected]
