A PDF can draw a character without providing assistive technology with the correct Unicode value. Unicode identifies characters such as å, 7 and %. Problems may appear during copying, searching or speech. Embedded fonts alone therefore do not prove correct text.
Start by examining export and character encoding. Fix the cause there when possible.
Sources: [1]
Copy a word containing accented characters into a plain text area. Repeat with a negative amount and a unit containing an exponent, such as m². Incorrect output gives a concrete clue: visual appearance and usable text do not agree.
Copying is a diagnostic check, however. A correctly copied word does not establish heading structure, reading order or screen-reader behaviour. Also search for the word and read it in the intended environment. Record the failing characters or combinations; they may reveal a shared export or font problem.
Ligatures, where letters such as f and i share one visual shape, deserve attention. Readers need the correct word, rather than an unexpected symbol or missing characters. Prefer fixing character mapping or export when that is the cause instead of accumulating disconnected local replacements.
ActualText can provide replacement text when the ordinary textual representation is insufficient. This differs from Alt, which explains the meaning of content such as an image. Do not place long ActualText strings across whole pages to hide broken structure.
ActualText should represent the content it replaces. Replacing a minus symbol should not add an interpretation of the whole table. Where the problem affects a small sequence, keep the replacement appropriately scoped. Large replacements can make searching, selection and reading less intelligible.
Alt answers a different question: what information or function should readers receive from an image? A chart alternative may describe a comparison while its individual labels still need correct character representation. Switching a field from Alt to ActualText does not clarify a confusing chart.
After the edit, check for duplicated or missing content and continued navigation through the surrounding structure. Record the changed object and reason so the next editor can understand the repair after a new export.
Sources: [1]
Before
The word office looks right but copies as o□ce because the fi ligature lacks a usable character mapping.
After
Repair the exported character mapping or use narrowly scoped replacement text where needed.
Why this works Readers should receive the same word, not a paraphrase. Test searching, copying and speech after repair.
Check minus signs, decimal separators, units and exponents. An invisible error in −5 or m² can change meaning. Expressions requiring mathematical structure need treatment as mathematical expressions; a line of ordinary text does not solve every formula.
Sources
- Tagged PDF Best Practice Guide: SyntaxPDF Association
- Well-Tagged PDF 1.0PDF Association
Examples are our own unless stated otherwise. These guides distinguish legal requirements, standards and practical advice. Passing an automated check is not a complete assessment of document accessibility.