axe-core has been maintained by Deque since 2015. It's the accessibility audit engine running under Google Lighthouse, Microsoft Accessibility Insights, WAVE and most SaaS tools — including ComplAudit. When a tool claims to "detect accessibility problems", there's a good chance axe-core is underneath. Here's why, and where its power stops.
Why it became the standard
Three reasons: a very broad, well-maintained rule catalogue, one of the lowest false-positive rates on the market, and an open-source licence that made it the default engine of the whole ecosystem. When a rule evolves in WCAG, axe-core follows quickly. Building your own engine no longer makes sense: better to rely on the one the whole industry audits and fixes.
What it detects well (≈ 30–40% of criteria)
- →Images without alt, forms without labels, insufficient contrast
- →Heading structure (h1-h6 hierarchy)
- →Invalid or redundant ARIA attributes
- →Empty links, buttons with no accessible text
What it does NOT detect (≈ 60–70% remaining)
- →The relevance of an alt — it knows there is one, not whether it really describes the image
- →The logical tab order through a custom page
- →Video captions, audio description, transcript
- →The real behaviour of a modal with a screen reader
// Why 30-40% is already huge
The 30 to 40% that are automatically detectable are also the ones that take the most time to find by hand: you have to audit every page, every image, every form. An automated audit does this bulk work in a few minutes, freeing the human budget for the 60% that genuinely need a brain and a screen reader.
The right way to use it
axe-core is a starting point, not a finish line. The effective method: let it sweep the whole site to handle the mechanical non-conformities in bulk, then focus the human time on what it doesn't see — the relevance of alternatives, end-to-end keyboard navigation, real screen-reader rendering. Automated first, human next: not one without the other.
Frequently asked questions
Does a 100% axe score mean my site is compliant?
No. A perfect axe score only means none of the automatable rules fail — that's 30 to 40% of the criteria. The remaining 60% can still be a problem. "Zero axe violations" isn't "RGAA compliant".
Does axe-core measure the RGAA or WCAG?
WCAG. The RGAA is the French assessment framework, which builds on WCAG but with its own method and its own tests. A good RGAA tool uses axe-core as the engine, then maps and supplements towards the RGAA criteria.
Why do so many tools give different results?
Because they don't enable the same rules or the same tags. Two tools both based on axe-core can diverge depending on their configuration and the URLs tested. Consistency comes from the method, not just the engine.
An audit that combines axe-core and the full RGAA mapping:
→ Run an audit