Guides › How to
How to make a scanned PDF accessible
Short answer: Run OCR (optical character recognition) to add real text behind the page images, check that names, numbers and dates were recognized correctly, then tag the structure (headings, lists, tables), set the title and language, mark the page image itself as an artifact, and validate. If the original is available as a digital file, recreating the PDF from it gives a better result than fixing the scan.
Why scans fail
A scanned page is one image. A screen reader finds no text, no headings and no reading order, so the document is silent. WCAG treats this as an image of text.
Steps
- Start from the best source. If the document exists as a Word file or can be retyped, do that instead. A clean digital document beats any scan fix.
- Rescan if needed. 300 dpi, straight, high contrast. OCR accuracy drops quickly on skewed or faint scans.
- Run OCR. In Acrobat Pro: Scan & OCR › Recognize Text. Free options include OCRmyPDF (open source).
- Proofread the text that matters. Names, dates, dollar amounts, motion results. A screen reader reads what OCR produced, including mistakes.
- Tag the structure. Headings, paragraphs, lists, tables, in reading order.
- Mark the page image as an artifact so the screen reader does not announce a full-page picture on top of the text.
- Set title and language, then validate with PAC or veraPDF.
What our tool does with scans
Pages without text get OCR automatically; pages that already have an old hidden OCR layer get fresh OCR, because old layers are often inaccurate. We then tag the page and mark the scan image as an artifact. In our tests, clean scans came out well; old, speckled scans produced stray characters from marks on the paper, so the report asks you to spot-check those pages.
Frequently asked questions
- Is a scanned PDF with OCR text accessible?
- Not by itself. OCR makes the text searchable, but the document still needs tags for structure and reading order, and the OCR text must be accurate.
- Our old scans already have a text layer but it is full of errors. What then?
- Re-run OCR with a better engine, or correct the text. We automatically replace old hidden OCR layers with fresh OCR.
- What about handwriting and signatures?
- OCR handles printed text far better than handwriting. Signatures usually need a short text description, such as 'Signed by the City Clerk'.