AI Accessibility Solutions for Faster Compliance

Many organizations need a formal Accessibility Conformance Report (ACR) for procurement and compliance. Our VPAT tool supports templates for WCAG with Section 508, EN 301 549, and international variants. It ingests structured defect data from your accessibility audit and maps these issues to the right WCAG criteria. The output is an editable Word or PDF report that your accessibility team can review and finalize.

On average, this removes around 30 percent of the effort involved in producing VPATs and ACRs, while keeping humans in charge of the final statement.

Our documents tool takes untagged or partially tagged PDFs and HTML and brings them closer to accessibility standards using AI tools for accessibility. MagicA11y analyses each file for page count, structure, and existing tags. It uses our custom logic to add headings, lists, tables, and image tags.

You find that the output is a remediated PDF or HTML, along with a report listing any items that still need manual review. For clean, simple documents, the automation can cover most of the structure. Complex layouts still get human remediation and testing, including screen reader checks.

For organizations with large EPUB backlists, MagicA11y can work directly with existing files. Our tool ingests legacy EPUBs or HTML bundles and checks headings, lists, tables, and structural markup to identify issues. The AI tool then applies rule-based fixes where they are safe to automate, but leaves items that require human judgment as clearly marked for manual review.

The end result is a remediated EPUB or HTML package plus an Excel report with a clear fixed vs needs review status. This is how we lift large backlists toward current expectations without recreating every title from scratch.

Math content is often the hardest part of accessibility work. MagicA11y includes a dedicated pipeline for it. It converts PDFs to HTML using specialist libraries and can detect equations and expressions in text and images. Our teams spot-check the MathML with real screen readers, especially for dense STEM titles.

This approach makes it possible to handle books with thousands of equations, not just a few sample pages, while still keeping humans involved for accuracy.

Closed captions are not enough for many accessibility standards. Audio description is increasingly required. MagicA11y’s video tool accepts your video as input. It uses multimodal AI to identify key visual scenes and draft descriptions of what is happening on screen. High-quality text-to-speech is then generated for audio description segments. These segments are embedded back into the video timeline.

With an interface for editors to adjust text and timing, and then export the final enhanced video, this is the perfect tool for large video libraries accessible at a quicker pace. Moreover, our AI-enhanced video accessibility tool supports over 15 languages, with options to translate descriptions. You receive a single, enhanced video file with audio description included, not raw clips that still need manual post-production.

Most alt text tools or alt text generators just describe what is visually in the image. That usually is not enough. With MagicA11y, images can be uploaded in bulk, often as ZIP files. When images come from a PDF or book, our team selects relevant surrounding text as context. The tool generates
context-aware, discipline and grade-appropriate alt text. Our accessibility experts review and then export the final alt text for delivery.

MagicA11y has already been used to deliver tens of thousands of alt texts for clients, at a significantly lower per-image generation cost than an entirely manual process, with human review available.

We work with established engines and a private LLM. MagicA11y orchestrates and extends them with our own logic and workflows rather than asking you to throw your stack away. PDFs, EPUBs, HTML, images, video, math, and courseware packages are all supported in one approach.

MagicA11y is already used by our teams in live projects, including math-heavy higher education titles and high-volume alt text work. We deliver between 40 to 45 percent effort reduction on repetitive tasks using AI for accessibility. If a project or client requires purely manual methods, we support that too.

We treat AI in accessibility with the same seriousness as accessibility itself. MagicA11y keeps AI safe by minimizing data being sent to language models. For clients who cannot use public AI, we can work with private models and private infrastructure so that content never reaches public endpoints. Tools run on secure cloud infrastructure. Specialist engines, such as math conversion services, can run in secure dedicated environments. If your policy does not allow AI use on certain projects, we switch those workflows to fully manual while still using MagicA11y where it is permitted.