How to tell a page is scanned
- You cannot select a line of text; the whole page highlights as one image.
- Searching for a word you can see on the page finds nothing.
- PacketReady's "Real text, not scanned images" check lists pages that contain images but no selectable text.
Some scans already have an invisible text layer added by the copier or scanning software. That makes the page searchable, but the text is usually untagged and may contain recognition errors, so it still needs the steps below.
What OCR does and does not do
Optical character recognition looks at the image of each page and adds the characters it recognizes as real text behind the image. W3C technique PDF7 describes it as the way to provide actual text for a scanned PDF.
OCR alone is not the finish line:
- It makes mistakes. Accuracy depends on scan quality. Faint copies, skewed pages, stamps over text, small print and handwriting all cause errors, and a misread digit in a budget or bid tabulation changes its meaning.
- It does not add structure. Recognized text still needs tags for headings, lists and tables, and a reading order.
- It does not describe images. Signatures, seals, photos and maps still need alternate text or a written description.
The Matterhorn Protocol lists both problems as PDF/UA failures that a person has to check: OCR text with significant errors (08-001) and OCR text that is not tagged (08-002) (PDF Association).
Best option: get the digital original
A scanned document is often a printout of a digital file. Asking for that file usually takes less effort than repairing the scan:
- Add a line to agenda submission instructions: "Please send documents as Word files or PDFs saved from the original program. Send scans only when no digital version exists."
- Ask departments to sign electronically, or to send the unsigned digital version along with the signed scan for your records.
- In contracts and requests for proposals, require deliverables that will be posted publicly to be accessible digital files.
When you must OCR
- Scan well. Straight, clean pages at the resolution your OCR software recommends for text. Remove sticky notes and avoid dark copier backgrounds.
- Recognize the text. In Acrobat Pro: All tools > Scan & OCR > Recognize text > In this file (Adobe help).
- Add tags. All tools > Prepare for accessibility > Automatically tag PDF, then fix headings, lists and tables that were misread.
- Proofread. Check names, dates, dollar amounts and vote counts against the image.
- Describe images. Give signatures, seals and photos alternate text, or mark purely decorative marks as artifacts.
- Finish the settings. Add a document title and language, then run the file through PacketReady again.
Signatures, handwriting, stamps and plans
- Signatures. Alt text such as "Signature of the Village President" is enough; the important part is that the text of the document itself is real, tagged text.
- Handwriting. OCR does not read handwriting reliably. Provide a typed transcript of handwritten comments, notes or forms.
- Stamps and seals. Transcribe any information that matters, such as a "Received" date or a notary's commission details.
- Large drawings and maps. OCR will not make a plan set accessible. Add a written summary of what the drawings show and a contact for questions.
Scans inside combined packets
A single scanned attachment can make an otherwise accessible packet fail. PacketReady reports which pages are affected, so you can fix or replace just those pages. If you post attachments as separate files, a scanned attachment does not hold up the rest of the packet while it is being fixed.
Check your own files
PacketReady checks a whole packet folder in your browser, with nothing uploaded, and explains every fix in plain English.
Related guides: Accessible board packets ยท Alt text for meeting documents