Are PDFs accessible to assistive technologies and AI?

This is a great question! And the answer is “it depends”:

PDFs are accessible to AI when:

•    The PDF contains real, selectable text (not just scanned images)
•    It has a logical reading order
•    Headings, lists, and tables are properly structured (tagged PDFs)
•    The file is not blocked by permissions or behind a login

In these cases, modern AI systems (including search engines and LLMs) can read, index, summarize, and reference the content — similar to how they process a web page. This is true for general ADA accessibility compliance as well.

PDFs are not accessible to AI when:

•    They are scanned images without OCR
•    Text is embedded as images
•    The structure is flat or broken (no headings, no tags)
•    The file is locked or inaccessible to crawlers

These PDFs are effectively invisible to AI (and often to assistive technologies as well).

Should you use PDFs when sharing content?

Relying on PDFs alone to share crucial information is risky, and can limit discoverability. Cybernautic's recommendation would be to always have the information within the PDF accessible on a standard webpage — at least the most important parts. You can then offer the PDF as a downloadable, printable resource on the page. 

In Summary:

PDFs can be read by AI, but only if they’re properly structured and text-based. However, AI systems (and search engines) work best with native web pages. The best practice is to publish key content as standard content on the website and offer PDFs as downloadable supplements, rather than relying on PDFs alone.