Creating an Open Source, Customizable Accessibility Checker for Content Authors
Adam Chaboryk · 2022 · Proceedings of the 19th International Web for All Conference (W4A) · doi:10.1145/3493612.3520468
Summary
This extended abstract introduces Sa11y, an open-source, in-page accessibility checker specifically designed for content authors rather than developers. Created at Ryerson University (now Toronto Metropolitan University) to manage accessibility across several hundred website editors and over 40,000 web pages, Sa11y addresses a practical gap in the accessibility tooling landscape. Enterprise accessibility suites are powerful but expensive and often overwhelming for non-technical users. Sa11y takes a different approach: it runs directly in the page, highlights errors and warnings at their source with simple tooltips explaining how to fix them, and covers over 50 test conditions spanning heading structure, alternative text, hyperlinks, tables, multimedia, and contrast. Built with vanilla JavaScript, CSS, and HTML, Sa11y is platform agnostic and designed for easy integration into any content management system. It is an adaptation of Khan Academy's Tota11y tool, extended with customization capabilities and a focus on role-based accessibility responsibilities.
Key findings
Sa11y's design philosophy centers on five principles. First, a role-based approach aligned with the W3C's Accessibility Roles and Responsibility Mapping (ARRM) framework ensures content authors only see issues they are responsible for — developers can configure Sa11y to check only editable content areas, avoiding erroneous warnings about navigation or other developer-owned components. Second, the tool emphasizes quality assurance beyond simple pass/fail checks; for example, it not only flags missing alt text but lets authors review all image descriptions for relevance and quality, including specialized guidance for functional images used as hyperlinks. Third, Sa11y is highly customizable — developers can create custom rulesets using simple JavaScript to enforce organization-specific usability guidelines, such as discouraging overuse of particular components. Fourth, false positives are minimized through CSS selector-based exclusions for elements like auto-populated avatars or navigation links. Fifth, the tool provides three clear states (error, warning, pass) giving non-technical authors positive affirmation when content is barrier-free.
Relevance
Sa11y represents an important shift in accessibility tooling philosophy: rather than building ever more comprehensive developer-focused scanners, it targets the people who actually create and maintain web content day-to-day. For organizations managing large websites with many content contributors, this approach distributes accessibility responsibility more effectively than relying solely on developer-side auditing. The role-based model, combined with customizable rulesets and false-positive filtering, makes automated accessibility checking practical for users without technical backgrounds. The tool's open-source nature and framework-agnostic design lower adoption barriers significantly. However, as a two-page extended abstract, the paper lacks formal evaluation data — there are no user studies, accuracy benchmarks, or comparative analyses with other tools. The practical impact at Ryerson is implied but not measured.
Tags: automated testing · accessibility tools · content management · open source · quality assurance · web accessibility
Standards referenced: WCAG · ARIA · ARRM