Adobe’s new AI automates PDF accessibility tags

Depending on who you ask, AI can be mesmerizing, terrifying or some combination of both. But Adobe’s accessibility-focused artificial intelligence is a use case that (I hope) any reasonable person can get behind. The company’s new Auto-Tag API can remove the tedium from making accessibility-friendly PDFs for people with disabilities.

PDFs have built-in metadata providing structural information — headings, paragraphs, lists and tables — in documents for assistive tech like screen readers. But tagging digital documents for compatibility can be time-consuming, especially for those with complex layouts or companies with backlogs of old files lacking the proper metadata. Adobe estimates that over 90 percent of PDFs are at least partially inaccessible for people with disabilities, sometimes “appearing blank, blurry or as lines of distorted text.”

Adobe’s PDF Accessibility Auto-Tag API automates the tagging process. The company says its Sensei-powered software will indicate the correct reading order for assistive technology, saving companies time and — more to the point — making PDFs more readable for people with disabilities. Adobe says the AI can quickly go through stockpiles of old documents lacking the proper structure.

The company is also launching a PDF Accessibility Checker, which will “enable organizations to quickly and easily evaluate the accessibility of existing PDFs at scale.”

In addition to the developer-facing API, Auto-Tag is coming to Acrobat Pro in June, offering automated metadata structuring within the app at no additional cost for all subscribers. The feature will also make its way to Acrobat Reader this fall.