Lighthouse Accessibility Benchmark

The Lighthouse accessibility benchmark measures how well a web page meets automated accessibility checks. Lighthouse, an open-source auditing tool built into Chromium developer tools and available in CI, runs a battery of accessibility audits derived largely from the axe-core engine and aligned with Web Content Accessibility Guidelines (WCAG). It produces a 0-100 score plus a list of specific failures.

Accessibility is both an ethical and often legal requirement, and automated scoring gives teams a fast, repeatable signal to catch regressions, though it covers only part of true accessibility.

What It Measures

The accessibility score is a weighted aggregate of individual audits, each of which passes or fails. Audits check for issues such as insufficient color contrast, missing alt text, improper ARIA usage, missing form labels, poor heading structure, and keyboard/focus problems. The benchmark also reports the count of passed and failed audits and lists each failing element, which is the actionable output.

Methodology

Lighthouse loads the page in a controlled browser environment and runs each accessibility audit against the rendered DOM. Most checks come from axe-core, which evaluates the accessibility tree, ARIA attributes, contrast ratios, and semantic structure. Each audit is weighted by impact, and the weighted pass/fail results combine into the 0-100 score. Because the audit runs on a single rendered state, dynamic content and authenticated flows require scripted navigation to a representative state before auditing. Running Lighthouse in CI on key pages catches regressions, and comparing scores across builds tracks progress. The detailed report attributes each failure to specific elements with remediation guidance.

How to Interpret Results

Treat the score as a floor, not a certificate of accessibility. A perfect 100 means automated checks passed, but automated tooling can only detect roughly a third to half of WCAG issues; many criteria require human judgment. The failing-audit list is more valuable than the number, since each entry is a concrete, fixable problem. Prioritize high-impact failures like missing labels and contrast, which affect many users. Use the score for regression detection across builds rather than as the sole accessibility goal, and always supplement it with manual testing, screen-reader checks, and keyboard navigation.

Limitations

Automated accessibility testing catches only a portion of real barriers; issues like meaningful focus order, sensible alt-text content, logical reading flow, and usable interactions with assistive technology need human evaluation. A high Lighthouse score can create false confidence. The audit runs on one page state, missing dynamic, authenticated, or multi-step flows unless scripted. Scores can fluctuate with content and viewport. The benchmark is best used as a fast regression guard and starting point, integrated into CI and paired with manual and assistive-technology testing for genuine conformance.