You answer for what is on the site. The obligation to make the PDFs on it work for human visitors was familiar before AI systems started reading the same pages. It is no longer the only obligation those documents carry.
The first obligation is already familiar
Accessibility law treats the PDFs on a public-facing website as in scope. In the United States, the Department of Justice Title II final rule takes effect in April 2026 for public entities serving populations of 50,000 or more, and in April 2027 for smaller entities. In the United Kingdom, the Public Sector Bodies Accessibility Regulations 2018 apply regardless of size. In the European Union, the European Accessibility Act applies from 28 June 2025.
Section508.gov and GOV.UK Government Digital Service both direct publishing toward HTML. PDFs are treated as the format to avoid or replace, not as the default to protect. The direction of travel is not ambiguous, and the enforcement calendar is live.
The second obligation is newer, and less well understood
AI readiness is the half most operators have not yet brought to the leadership table. It describes whether the organisation's external digital presence can be read, interpreted, trusted, and acted on by AI systems, and whether the organisation still controls what that presence says about it.
The evidence that this is live is not speculative. Google has expanded AI Overviews to more than 200 countries and 40 languages, with usage growth above 10% in major markets. Adobe reports AI-sourced traffic to US retail sites rose 393% year on year in the first quarter of 2026, with stronger engagement than non-AI traffic. Deloitte reports 9 in 10 retail executives expect AI to be used more than search engines by 2026.
OpenAI has expanded shopping and product discovery inside ChatGPT, with structured merchant feeds governing discoverability, relevance, and trust. The US National Institute of Standards and Technology Generative AI Profile identifies provenance, misinformation, and tampered content as risks that require active identification and mitigation. Taken together, this is not an emerging layer. It is an operating one.
The known online, and the unknown online
The distinction that governs the PDF estate is the one between the known online and the unknown online. The known is what the organisation actively manages. The unknown is what is still visible externally but no longer actively governed, and it almost always includes PDFs published years ago and left in place on pages no one has reviewed recently.
AAAnow / Sitemorse internal analysis of 100+ million websites indicates 41% of websites are typically unknown to the organisation's own digital teams. The same internal analysis indicates 19% of PDFs on organisational websites are duplicates. Both figures are internal data, labelled as such, and they are directional in what they measure. The point they support is that most organisations know less about their own estate than they assume.
Where the two halves meet
The two halves of the obligation meet on the same documents. A PDF that a screen reader cannot parse is structurally opaque to a retrieval system as well. Duplicates and stale content that confuse a visitor are also provenance weaknesses for language models. A document left in place for years has aged out of its own guidance, and has continued to feed the picture external systems form about the organisation that published it.
This is why the PDF estate now sits inside two obligations at once. Inclusion is about the people in front of the screen. AI readiness is about the systems reading the same content behind it. Neither is optional, and neither can be resolved by work that addresses only the other.
What to do about it
Recognising the obligation is only the first move. The operational stages that follow, discovery, triage, prioritisation, execution, and version management, are covered in The five stages of dealing with a PDF estate. The options the market offers as responses to those stages, and an idea of what they cost to use, are covered in The PDF market, remediation options and idea of costs. What happens to the content once it has left the site, and why that matters for product catalogues, recalls, and distribution through partners, is covered in What happens when a PDF leaves your site.