The Accessible PDF With Hidden Structure Problems
You must publish a 40-page annual report as a PDF for public release. It contains a cover image with the organisation name baked into the image, 12 charts, a two-column section, and a footer with page numbers. The document was built with headings for visual effect only - several bold paragraphs look like headings but are not tagged. A colleague exported it and the resulting PDF has no selectable text in the two-column section and reads the footer before the body on every page.
1. Explain the two symptoms (unselectable text, footer read first) and what each indicates about the export.
Read the worked answer
Unselectable text in the two-column section indicates the content was exported as an image rather than as text - typically a section that was pasted or rendered as a picture, or an export setting that rasterises complex layout. Selectable text requires real text objects in the PDF.
Footer read before the body indicates the reading order / tag order is wrong: the footer was placed earlier in the content stream or tag tree than the body, so assistive technology and text extraction encounter it first. This is a tagging/structure problem, not a visual one - the page looks fine.
Both are structural export failures, invisible on screen, which is why the colleague's visual check passed.
2. The bold paragraphs that look like headings are not tagged. Why does this matter for the PDF, and what is the fix?
Read the worked answer
In a PDF, headings become navigation and structure (bookmarks, tag hierarchy) only if the source paragraphs carry heading styles. Bold paragraphs carry no such meaning, so the exported PDF has no heading structure: screen-reader users cannot jump between sections, and the document outline is empty or wrong.
The fix is at the source, not the PDF: apply real heading styles to those paragraphs, then re-export. Tagging cannot be retrofitted reliably by hand in a 40-page file. This is the same lesson as the TOC - visual appearance is not structure.
3. The cover has the organisation name inside the image. What must be done before release, and how does this differ from the charts?
Read the worked answer
The cover image's baked-in text is not accessible - it cannot be read by assistive technology. Add alternative text describing the organisation name and any essential information in the image, or better, place the organisation name as real text over a decorative image and mark the image decorative.
The charts differ: each needs alt text that conveys the finding, not just "a chart" - e.g. the trend and the key figure - because the information is in the data, not the picture. A generic "chart of revenue" is technically present but useless.
The distinction: cover text needs alt text to convey identity; charts need alt text to convey meaning. Both are required, but a single template alt text would fail the charts.