Your Section 504 deadline moved. A lot of guidance still says it didn’t.
HHS extended the Section 504 web accessibility compliance dates by one year in an interim final rule published on May 11, 2026 — the same day the original deadline fell. Recipients with 15 or more employees now have until May 11, 2027, roughly 7 months away. The requirement is WCAG 2.1 Level AA and it covers PDFs.
The current dates, with their sources
| Who | Deadline | Was | Rule |
|---|---|---|---|
| Recipients of HHS funding with 15 or more employees — hospitals, health systems, most clinics | May 11, 2027 | 11 May 2026 | HHS interim final rule, 11 May 2026 (FR doc. 2026-09266) |
| Recipients of HHS funding with fewer than 15 employees | May 10, 2028 | 10 May 2027 | HHS interim final rule, 11 May 2026 (FR doc. 2026-09266) |
Two cautions about anything you read on this, including this page. First, a great deal of published commentary predates the extension and still says May 2026, because the rule landed on the deadline itself. Second, reputable secondary sources currently disagree about the small-recipient date — we have seen both 10 and 11 May 2028 in print. The Federal Register says 10 May 2028, and that is what is above. Check the rule, not a summary of it, and that includes checking ours.
An extension is not a reprieve from the obligation, and it does not affect existing Section 504 duties or private litigation risk. It moves one date.
Why PDFs are the part that gets missed
Website accessibility work almost always starts with the website: colour contrast, focus order, form labels, alt text on images in the CMS. That work is visible, it has obvious owners, and there are good tools for it.
The PDFs are different. They were produced over fifteen years by different departments in different software, they sit behind links rather than in the CMS, nobody owns them collectively, and there is usually no list of them anywhere. A patient rights notice, a financial assistance application, a discharge instruction sheet, a price transparency file, a plan summary — each one is a document somebody has to be able to read, and each one was exported from Word by a person who had no reason to think about tagging.
So an organisation can complete a full website accessibility project and still have several thousand inaccessible documents linked from it.
What “inaccessible” means to a patient
Not an abstraction. A PDF with no tag structure is read by a screen reader as one unbroken block of text — no headings to jump between, no list boundaries, no table rows. A form with unlabelled fields is announced as a series of empty edit boxes, so it cannot be completed without sighted help. A scanned document has nothing to read at all: the screen reader reaches it and falls silent, and nothing announces that anything is missing.
These are not edge cases in medical publishing. Financial assistance applications are forms. Lab result summaries are tables. Older policy documents are scans.
How common is this? Here is a measurement
We scanned 15 published federal documents on September 27, 2026 — tax forms, a statute, a CDC statistical report, a Federal Reserve publication. All are public, and all come from well-resourced organisations with accessibility obligations of their own.
All 15 of 15 failed PDF/UA conformance. 3 have no tag structure at all. The published text of the ADA Title II web accessibility regulation is itself among the documents that fail.
That is not here as a gotcha. It sets a realistic expectation: if you scan your own portfolio and most of it fails, you have a normal document estate, not a negligent one. Nobody’s portfolio is clean. The useful question is not whether you have failures but which of them actually stop someone using a document, and what fixing those costs.
What a scan tells you that a checklist does not
| Question | Why it decides your budget |
|---|---|
| How many PDFs do we actually have? | Almost nobody knows. The number is usually several times the estimate, and it is the multiplier on every cost figure you will produce. |
| How many are scans with no text layer? | The expensive ones. A scan has to be rebuilt rather than corrected, and it sits at the top of the $5–$25 per page range. |
| Which failures stop a document being usable? | A missing document title and an unlabelled form field are both failures. One takes seconds; the other means the form cannot be completed. |
| Which documents do people actually open? | A blocker on a document linked from your homepage matters more than the same blocker on a 2014 board minute. Crawl depth and inbound links approximate this. |
What this costs
Remediation — repairing a document’s tags — is quoted publicly by established vendors at roughly $5 to $25 per page. The spread is real: a tagged document missing a title needs a minute, and a scan needs rebuilding from scratch.
That range is why triage comes first. At $25 a page, a two-thousand-document estate averaging ten pages is a seven-figure number if you treat every document as equal. Most are not equal, and a large share of most estates does not need remediating at all — documents that are obsolete, superseded, duplicated across three URLs, or that should have been a web page in the first place.
Put a range around your own estate with the cost calculator before you talk to a remediation vendor. Knowing the number is what makes that conversation a scoping exercise rather than an open-ended one.
What we do, and what we deliberately do not
PDFCraft does triage. Point it at a domain; it finds the PDFs, checks each one, and returns a prioritised report: what fails, how severely, what it would cost, and what to fix first. We do not remediate documents. Fixing PDFs is a real service business with established vendors, and we are the step before you call one — the step that tells you which documents to send them.
The first 25 documents are free and need no account, because the report is what makes the case internally and it is no use to us behind a form.
What an automated scan cannot do
It cannot certify a document as accessible. The PDF Association’s Matterhorn Protocol divides its checkpoints into machine-verifiable and human-judgment categories, and no tool clears the second set. What a scan does well is the opposite job: finding, at scale, the documents that definitely fail. For a portfolio of thousands that is the constraint that matters — you cannot hand-review five thousand PDFs, and you do not need to, because a large fraction fail on something a machine can see.
Our checks go one step beyond a conformance validator: we compare the reading order stored in the file against the visual order on the page, test whether something tagged as a table is really a grid, and flag alt text that is a filename rather than a description. Those are judgments a validator does not make, and they are where documents that “pass” still fail a real reader. The overview explains how.
Related
- All four current deadlines, with Federal Register citations
- PDF remediation cost calculator
- Scanned PDFs: the expensive category
- PDF/UA and WCAG 2.1 AA: how they relate
Deadlines verified September 27, 2026 against the Federal Register. Scan performed September 27, 2026 with pdfa11y 0.0.11 plus PDFCraft’s geometric checks. This page is not legal advice.