Translate Arabic and English PDFs Without Layout Errors
Answer: You can translate Arabic and English PDFs without major layout errors if you OCR scans first, use a layout-preserving PDF translator, and proofread the output. Test a one-page sample, check numbers and mixed Arabic-English lines, and fix the remaining layout issues by hand.
Why layouts break when you translate PDFs
PDFs store visual layout, not always editable text. Translators usually work on strings, not on page boxes. Arabic adds two problems. Arabic reads right to left. English and numbers read left to right. If your tool ignores bidirectional rules, words or product codes reverse. Scanned PDFs add a third problem. The translator must run OCR first and then map text back into the layout boxes.
Independent tests show many tools still fail on complex pages. TryReflo measured multi-column accuracy at about 38.4 percent for Google Translate in their 2026 benchmark. That means tools often change column order or split paragraphs on multi-column pages. See Sources for the full report.
Quick checklist before you start
- Check if the PDF contains selectable text or just images. Selectable text = faster path. Image-only pages need OCR.
- Make a copy of the original PDF. Work on the copy.
- Test on one representative page. Use that to compare layout fidelity.
- Identify critical elements that must stay exact: numbers, dates, product codes, tables, legal stamps.
- Note file size and page count. Google Translate web UI limits are 10 MB and 300 pages. Use a cloud API for larger files.
Tools and limits you should know
- Google Translate web supports PDF uploads up to 10 MB and 300 pages. It often preserves simple layouts but struggles with scans and complex multi-column pages. Source: Google Translate help.
- DeepL supports PDFs and often keeps layout well for simple files, but file size limits depend on your plan. Check DeepL support for plan specifics.
- Adobe Acrobat AI Assistant can translate inside Acrobat and preserve layout in many cases. Adobe added these features in 2024 and later. Check Adobe for current availability and pricing.
- Enterprise document translators and some PDF editors advertise better layout preservation. Independent benchmarks vary, so test with your files. TryReflo published a 2026 benchmark comparing nine tools across 240 documents.
- For scanned PDFs, run OCR first. Use a reliable OCR engine set to Arabic and English to reduce errors. FreeToolr offers a simple OCR PDF tool at /pdf/ocr-pdf/ you can use before translation.
Step-by-step workflow
A. If your PDF has selectable text
- Make a copy of the PDF.
- Export or upload to a layout-preserving translator or to a PDF editor with translation features. Prefer tools that state they keep layout and support bidirectional languages.
- Translate the file. If the tool has a preview, compare the original and translated pages.
- Proofread numbers, codes, and table alignment. Fix reversed code fragments and misplaced punctuation.
- Save as PDF. If fonts changed, embed fonts to avoid substitution.
B. If your PDF is scanned or images-only
- Run OCR with Arabic and English selected. If you use FreeToolr, try /pdf/ocr-pdf/. OCR turns images into selectable text you can translate.
- Export the OCR result as a searchable PDF or as Word if your translator prefers Word input.
- Use a layout-preserving translator that accepts searchable PDFs or Word.
- Review bidirectional lines. Pay special attention to numbers, dates, ISBNs, product SKUs, and email addresses.
- Correct tables and multi-column flow manually when the automatic result breaks columns or cells.
How to handle Arabic and English mixed text
You must preserve direction for mixed lines. Use these checks and settings:
- Set OCR and translator to recognize Arabic and English.
- Use a tool that supports Unicode bidirectional rendering.
- After translation, proofread each line that mixes Arabic words and LTR substrings.
- For critical numeric strings, mark them as "do not translate" or place nonbreaking markers in the source before automatic translation. If the tool lacks that option, check every numeric field manually.
Example problem: a product code like ABC-123 must remain ABC-123 and not become 321-CBA. Test codes on a sample page.
Tables, formulas and multi-column pages
Tables and formulas are frequent failure points. Automatic translators may reflow table cells, drop lines, or disrupt column order. Use this strategy:
- Isolate complex tables and export them to Word or Excel. Translate cell by cell if needed.
- Reinsert translated tables into your PDF editor and adjust cell width or font size to fit.
- For formulas or code, do not machine translate them. Keep original images or ask a human translator with subject knowledge to verify.
Fonts and embedded characters
PDF fonts can break after translation. If a translated phrase uses letters not present in the original font, the viewer will substitute another font. Always embed fonts when you save the translated PDF. If your output looks wrong in one PDF viewer, test it in Acrobat Reader and another viewer.
Proofreading checklist for Arabic-English PDFs
- Check numbers, dates, and IDs for correct left-to-right order.
- Verify column order on multi-column pages.
- Open side-by-side with the original and confirm paragraph breaks.
- Confirm table cell alignment and missing text.
- Verify embedded images, charts and captions.
- Confirm fonts display correctly on other machines.
Worked example with times and sizes
You have a 42-page scanned user manual in Arabic. File size is 18 MB. It has two-column pages and product codes mixed with Arabic. Here is a realistic workflow and time estimate:
- Make a copy and split the document if needed. Splitting reduces upload size.
- Run OCR on all pages set to Arabic plus English. Expect OCR to take 8 to 12 minutes on a typical modern laptop or cloud OCR instance for 42 pages.
- Export the OCR result as a searchable PDF. File size after OCR stays near 18 MB.
- Upload the searchable PDF to a layout-preserving translator or to Acrobat AI Assistant. If you use Google Translate web, note the 10 MB limit, so you must use another tool or split the file.
- Auto-translate pages. Translation time depends on service speed. Expect 2 to 10 minutes for 42 pages with a cloud translator.
- Spot-check five representative pages for column order, product codes, and tables. This should take 20 to 30 minutes.
- Manually fix table cells and mixed lines. Allocate about 60 to 90 minutes for manual fixes for a document of this complexity.
- Final review and font embedding: 10 to 20 minutes. Total working time estimate: about 2.5 to 4 hours. Your time will vary by tool, internet speed, and how many manual fixes you need.
Privacy and legal notes
If your PDF contains personal data, check the translator vendor's privacy terms and data residency options for GDPR compliance in the EU and UK. For sensitive or legal documents, use a paid business plan that offers non-retention guarantees or use on-premise tools. DeepL and other vendors publish security pages and SOC notices you can review.
For US immigration and many legal filings, machine translation alone does not meet requirements. The US Code of Federal Regulations 8 CFR 103.2(b)(3) requires a certified human translation for documents submitted to USCIS. Use a certified translator for filings and official submissions.
When to call a human translator
Use machine translation for reading, quick drafts, and internal use. Use a certified human translator when you must file documents with government agencies, courts, or for legally binding contracts. Use a human for highly technical texts, legal terms, or where layout fidelity affects meaning.
Tools recap and how to test them
- If the PDF is text-based, try a layout-preserving translator or Acrobat AI Assistant.
- If the PDF is scanned, OCR first with Arabic and English. Try FreeToolr OCR at /pdf/ocr-pdf/ before translating.
- Test one page first and compare originals.
- Check specific problem areas like tables, formulas, and mixed direction lines.
Sources
- Google Translate upload limits: https://support.google.com/translate/answer/2534559
- Google Cloud Translation documents guide: https://cloud.google.com/translate/docs/advanced/translate-documents
- DeepL file formats and plans: https://support.deepl.com/hc/en-us/articles/360020582359-File-formats
- Adobe Acrobat translate resources: https://www.adobe.com/acrobat/resources/translate-pdf.html
- TryReflo 2026 layout preservation benchmark: https://blog.tryreflo.com/2026-pdf-translation-format-preservation-benchmark-report-reflo-vs-9-leading-tools-tested-across-240-real-documents/
- DeepReference guide on Arabic PDF translation: https://www.deepreference.com/blog/translate-pdf-english-to-arabic
- USCIS certified translation rule (8 CFR 103.2): https://www.law.cornell.edu/cfr/text/8/103.2
- EMNLP 2025 PDFMathTranslate demo: https://aclanthology.org/anthology-files/pdf/emnlp/2025.emnlp-demos.71.pdf
- ForMaT and BabelDOC research: https://arxiv.org/abs/2605.15794 and https://arxiv.org/abs/2605.10845
Questions people ask
Can I translate scanned Arabic PDFs automatically?
Yes, but you must run OCR first. Set OCR to Arabic and English and export a searchable PDF. Then use a layout-preserving translator and proofread the result.
Will Google Translate keep my PDF layout intact?
Sometimes for simple text PDFs. Google Translate web limits uploads to 10 MB and 300 pages. It often fails on scans, multi-column pages, and complex tables.
How do I keep numbers and product codes from reversing in Arabic text?
Use a translator that supports bidirectional text and proofread every mixed line. For critical codes, mark them as do-not-translate or check them manually after translation.
Is machine translation acceptable for immigration documents?
No. USCIS requires a certified human translation under 8 CFR 103.2(b)(3). Use a human certified translator for filings and official submissions.
Which tool should I try for OCR before translating?
Use a reliable OCR engine that supports Arabic. FreeToolr offers an OCR PDF tool at /pdf/ocr-pdf/ you can use to convert scans to searchable PDFs.
FreeToolr articles are written and checked by our editors. Found a mistake? Email [email protected]. Read our editorial policy.