Comparison
axe-core vs WAVE
Two of the most widely used free accessibility checkers, compared - a rules engine for automated tests versus a visual tool for reviewing pages.
Published
Quick summary
Use axe-core when you want accessibility checks to run automatically in code and CI; use WAVE when a person wants to see problems on a page and understand them. They are complementary rather than rivals, and many teams use both - axe-core in the pipeline and WAVE for design, content and manual reviews.
Key takeaways
- axe-core is a JavaScript rules engine built into tests; WAVE is a visual evaluation tool.
- Both are free to start with; WAVE charges for its APIs.
- axe-core suits developers and CI; WAVE suits designers, editors and auditors.
- Neither can confirm full WCAG conformance without manual testing.
Side-by-side features
| Feature | axe-core | WAVE |
|---|---|---|
| Type of tool | Rules engine (JavaScript library) | Visual evaluation tool (website and browser extension) |
| Made by | Deque Systems | WebAIM |
| Cost | Free, open source (MPL 2.0) | Free website and extensions; paid APIs |
| Standards covered | WCAG 2.0, 2.1 and 2.2, levels A, AA and AAA, plus best practices | WCAG errors, contrast and alerts |
| Runs in CI | Yes, through test frameworks | Through the paid API |
| Shows issues on the page | Not by itself; through tools built on it | Yes, with icons in context |
| Pages behind a login | Yes, inside your tests | Yes, with the browser extension |
| Best known for | Low false positives; powers Lighthouse's accessibility audits | In-context visual feedback; engine behind the WebAIM Million study |
Pricing
axe-core
Free and open source. Deque sells commercial products built on it.
WAVE
The WAVE website and browser extensions are free. The subscription API and stand-alone engine are paid.
Accessibility testing
axe-core
Automated checks against WCAG 2.0 to 2.2. Deque says it finds on average 57% of WCAG issues; uncertain results are flagged for manual review.
WAVE
Flags errors, contrast failures and alerts, and shows page structure and ARIA to support a human review.
Security testing
axe-core
Not a security tool. Runs locally in your own test environment.
WAVE
Not a security tool. The extensions evaluate pages inside your browser.
Reporting
axe-core
Structured JSON results (violations, passes, incomplete) for your test reports.
WAVE
Visual icons on the page plus a summary panel; the API returns structured data.
Automation
axe-core
Built for automation. Call axe.run() from unit, integration or end-to-end tests.
WAVE
Mainly manual. Automation needs the paid API or a platform built on WAVE.
AI readiness
axe-core
Machine-readable output is easy for scripts and AI coding assistants to act on.
WAVE
Not designed for automated or agent use in its free form.
Best for
axe-core
- Developers adding accessibility checks to CI
- Catching regressions on every build
- Teams building their own tooling
WAVE
- Designers and content editors checking their pages
- Manual accessibility audits
- Learning how accessibility problems look on a page
How they differ in practice
axe-core lives in your code. A developer adds it to the test suite once, and from then on every build checks for detectable accessibility failures. Nobody needs to remember to run it.
WAVE lives in the browser. Anyone on the team can open a page, run the extension and see exactly which image lacks alternative text or which field has no label. It is a learning tool as much as a testing tool.
Because they work at different stages, they rarely compete. A common setup is axe-core in CI to stop regressions, WAVE for design and content reviews, and manual keyboard and screen reader testing before each release.
Verdict
Choose axe-core for automated, repeatable checks in code, and WAVE for visual, human review. If you can, use both.
Frequently asked questions
Does WAVE use axe-core?
No. WAVE uses WebAIM's own evaluation engine. axe-core is Deque's engine, used by tools such as Lighthouse.
Which finds more issues, axe-core or WAVE?
They use different rules and are designed for different jobs, so results differ. Running both, plus manual testing, gives the fullest picture.