Skip to main content

Why is the text in my PDF garbled?

Garbled PDF text usually points to one of three problems: the viewer cannot render the intended font, the PDF lacks a reliable character-to-Unicode mapping, or optical character recognition has produced an inaccurate text layer. Start by checking whether the characters look wrong on the page or only become gibberish when you select, search, copy, or convert them; that distinction identifies the right path for this PDF editing and text problem.

Visible text and extracted text rely on different PDF information

What the PDF stores (same for all three) U+0043 U+0061 U+0066 U+00E9 · U+0072 U+00E9 U+0073 U+0075 U+006D U+00E9 = the characters C a f é  r é s u m é
The same stored text, rendered under three conditions
Condition Font Encoding map What you see
Correct Embedded, present Matches text Café résumé
Font missing Not embedded, no substitute glyph Matches text □□□□ □□□□□□
Wrong encoding Present Mismatched (bytes read as Latin‑1) Café résumé
Garbled text isn't a typo in the file — the stored characters are fine; the font or encoding layer that turns them into shapes is missing or mismatched.

A PDF does not necessarily store text in the same straightforward form as a word-processing document. It can store character codes, positioning instructions, font resources, and mappings that tell software which glyphs to draw and which Unicode characters those glyphs represent.

A glyph is the visible shape of a character. Unicode identifies the character itself, such as a Latin capital A, a Greek alpha, or a particular punctuation mark. A PDF can contain enough information to draw the correct glyphs while still lacking enough information to identify those glyphs correctly during copying or conversion.

That is why two PDFs with identical-looking pages can behave very differently when you search them, copy a paragraph, use a screen reader, or convert them to Word.

Font and rendering problems change what you see

If the text is already garbled while you are viewing the page, the problem is usually in the rendering path. Common symptoms include boxes, bullets, blank spaces, incorrect symbols, overlapping letters, or characters from the wrong writing system.

A PDF normally refers to one or more fonts. A font may be embedded in full, embedded as a subset containing only the required glyphs, or left unembedded. When the required font data is unavailable, the PDF viewer may substitute another font.

Substitution is not automatically a failure. Standard Latin text often remains readable because the replacement font contains equivalent characters. Problems become more likely when the original uses a custom encoding, a symbol font, an unusual script, specialist notation, or a font whose internal character information is incomplete. The replacement may not contain matching glyphs or may interpret the stored character codes differently.

An embedded font can also be defective. For example, a PDF may contain only part of the required font data, inconsistent width information, or an incorrectly constructed encoding. In that case, the file can fail even though its document properties describe the font as embedded.

If the PDF looks correct in one viewer but wrong in another, the file may depend on viewer-specific font handling. If it looks wrong in every viewer and on every device, the PDF itself is more likely to contain incomplete or invalid font information.

Character mappings control copying, searching, and conversion

A PDF can display perfectly while producing gibberish such as D#x7, unrelated accented letters, empty boxes, or random symbols when you copy its text. This usually means the visible glyphs are available but the software cannot determine which characters they represent.

PDF fonts can include a ToUnicode character map. This map connects the character codes used inside the PDF to Unicode values that other applications can understand. When a reliable map is present, a viewer can identify the visible glyph as the intended letter, number, punctuation mark, or symbol.

If the map is absent, incomplete, or wrong, extraction software may have to infer the characters from the font encoding and glyph names. That guess can work for ordinary fonts but fail for custom, subsetted, symbolic, or incorrectly generated fonts. A code that draws the shape of an A, for example, does not have to be stored as the normal Unicode value for A.

This produces several recognizable failure patterns:

  • Every copied character changes, but spaces and line breaks remain roughly correct.
  • Only one font, language, or page produces gibberish.
  • Letters copy as symbols even though they look normal in the PDF.
  • Search cannot find a word that is visibly present.
  • A PDF-to-Word conversion preserves the layout but inserts incorrect characters.
  • Ligatures such as fi, ff, or fl disappear, split incorrectly, or become unrelated symbols.
  • Text copied from a table arrives in an unexpected order as well as containing wrong characters.

Wrong character values and wrong reading order are related extraction problems, but they are not identical. Recognizable words arriving in the wrong sequence usually indicate that the page’s text objects or reading order do not match its visual layout. Completely unrelated characters point more strongly to an encoding or Unicode-mapping failure.

OCR errors affect scanned or image-based pages

A scan can look like a page of text while containing only a page image. To make that document searchable or selectable, optical character recognition creates a text layer based on the shapes it detects in the image.

OCR can confuse similar shapes, especially in a low-resolution, skewed, faint, marked, or compressed scan. Common substitutions include 0 for O, 1 for l, rn for m, and incorrect punctuation. The results can be much worse when the wrong recognition language is selected or the page uses handwriting, decorative fonts, mathematical notation, or multiple scripts.

Some scanned PDFs place the recognized text invisibly behind the page image. The page therefore looks correct because you are seeing the original scan, while copied text contains OCR mistakes. Other files replace or cover parts of the image with recognized text, making the errors visible as well.

OCR is also a common reason that only certain pages fail. A document may combine digitally created pages, clean scans, and poor scans in the same PDF, with a different text quality on each group.

Tests that separate fonts, encoding, OCR, and viewer failures

Do not begin by repeatedly converting or editing the file. First identify which part of the PDF is failing. A few controlled checks can distinguish a display problem from an extraction problem without changing the original.

What happens Most likely area Most useful first check
Text looks wrong and copies wrong Font, encoding, or damaged PDF content Open the downloaded file in a second PDF viewer
Text looks correct but copies as unrelated characters Missing or incorrect Unicode mapping Paste a short sample into a plain-text editor
Text looks correct but copied words contain plausible spelling errors OCR text layer Zoom in and check whether the page is a scan
Text works in one viewer but not another Viewer rendering or font-handling difference Compare the same downloaded file in two current viewers
Only one font or section fails Problem font or font-specific mapping Check which pages and text styles are affected
Words are correct but arrive in the wrong order Reading order or page construction Copy one simple paragraph instead of a whole column or table
Text cannot be selected at all Image-only page, outlined text, or copying restriction Try search, selection, and document security checks separately
Boxes or bullets replace visible characters Missing or unsupported glyphs Inspect the PDF’s font list and embedding status

Use a short, distinctive line for testing rather than copying an entire page. Include uppercase and lowercase letters, numbers, punctuation, and any characters that visibly fail. Paste the sample into a plain-text editor so formatting and font substitution in the destination application do not hide the result.

Then work through these checks:

  1. Save the PDF locally. A browser preview may use a different rendering engine from a desktop PDF application. Testing the downloaded file prevents a browser-preview problem from being mistaken for damaged content.
  2. Open the same file in another viewer. If only one viewer shows garbled characters, update or reset that application before rebuilding the document. If every viewer fails in the same place, concentrate on the PDF.
  3. Compare viewing with copying. If the page looks correct but the pasted sample is wrong, the glyphs can be drawn but the character identities are unreliable. Reinstalling fonts on your computer is unlikely to repair that extraction mapping.
  4. Search for a visible word. A failed search supports an extraction or OCR diagnosis, particularly when the word is clearly readable on the page.
  5. Check whether the page is a scan. Select individual words and zoom in closely. A scanned page often reveals image pixels, while an invisible OCR layer may produce a selection highlight that does not align perfectly with the printed characters.
  6. Inspect the Fonts tab in the PDF’s document properties. Labels such as Embedded or Embedded Subset are useful evidence, although they do not prove that every font mapping is correct. An unembedded problem font makes substitution a stronger possibility.
  7. Test more than one page. A failure limited to a title, formula, signature, table, language, or page range often identifies the exact font or OCR batch responsible.
  8. Keep the original unchanged. Perform OCR, conversion, editing, or PDF repair on a copy. Some fixes replace text, flatten content, or change the document’s layout and accessibility structure.

Repair the layer that is actually failing

The best repair is usually to return to the document that created the PDF and export it again. A source document still contains the intended characters and text structure, while a damaged PDF may contain only glyph codes that cannot be mapped back with certainty.

When the source file is unavailable, the right fallback depends on whether the visible page, extracted text, or OCR layer is defective.

Re-export PDFs with visible font or rendering damage

For text that already looks wrong on the page, open the original Word, design, publishing, or specialist application and create a new PDF with the required fonts embedded. Direct PDF export generally preserves text and font information more reliably than printing the document to a virtual printer or scanning a printed copy.

Before exporting:

  • Replace any unavailable or damaged font with a stable font that supports every required character.
  • Confirm that the font’s licence permits embedding.
  • Check non-Latin text, mathematical symbols, currency marks, ligatures, and specialist notation.
  • Use the application’s PDF export or save function rather than taking screenshots of the pages.
  • Review the exported PDF on a computer that does not have the source fonts installed.
  • Test copying and searching as well as visual appearance.

If only one viewer fails, first update it and download the PDF again. A fresh download can also eliminate incomplete browser transfers or cached previews. Rebuilding the document should be reserved for failures that follow the file across viewers.

Recreate or recover text when copy and paste produces gibberish

If the PDF looks correct but copied text is wrong, the file needs a better mapping between its stored character codes and Unicode. Re-exporting from the source application is the most reliable way to create that mapping because the source still knows the actual characters.

When you do not have the source, convert a copy of the PDF to an editable format such as Word and compare the result carefully with the visible page. Conversion software may recover the text when the PDF contains usable encoding clues, but it cannot guarantee the intended characters when those clues are missing or contradictory.

Check names, account numbers, dates, legal clauses, formulas, accented letters, and punctuation individually. A conversion can appear mostly correct while silently changing a few high-consequence characters.

If ordinary conversion repeatedly produces gibberish, OCR can be used as a fallback even for a digitally created PDF. In that workflow, the software recognizes the visible shapes rather than trusting the broken text mapping. OCR may recover readable text, but it can introduce recognition errors and lose some layout, links, tags, form behaviour, or typography. Run it on a copy and proofread the result against the visible page.

For a document that must remain visually unchanged, keep the original PDF as the reference and treat the converted or OCR-processed version as a working copy. Do not claim that the repaired copy is identical until its text has been checked.

Rerun OCR for scanned pages with incorrect text

For a scanned PDF, improve the page image before recognizing it again. Straight pages, clear contrast, sufficient resolution, and the correct recognition language all increase the chance of accurate text.

A practical OCR repair process is:

  1. Make a copy of the original PDF.
  2. Remove or replace the inaccurate OCR text layer if the software supports that operation.
  3. Correct page rotation and severe skew.
  4. Rescan pages that are blurred, clipped, shadowed, or heavily marked when the paper originals are available.
  5. Select the correct language or languages for recognition.
  6. Run OCR on the affected pages.
  7. Review uncertain words and compare the new text with the page image.
  8. Test search, copy and paste, and conversion on representative pages.
  9. Save the corrected result as a new file rather than overwriting the only original.

If the document contains tables, columns, handwriting, equations, or mixed languages, expect more manual correction. OCR is an interpretation of the page image, not a recovery of the original electronic text.

A PDF-to-Word conversion is a useful next step when you need an editable copy and the visible PDF remains readable. The converted file should still be compared with the original before it is reused, quoted, signed, or distributed.

Source: PDF character encoding, Unicode mapping, and text-extraction behaviour from PDF Reference, version 1.6, Adobe; font embedding and substitution guidance from Font embedding and substitution in PDFs with Distiller, Adobe; and OCR correction guidance from Fix text recognition errors in scanned PDFs, Adobe. Page last checked 2 October 2026.