If you want to translate a PDF to English, first check whether you can select the words in the document. If you can, the quickest route is usually to open Google Translate on a computer, choose Documents, upload the PDF, select English, and download the result. If you cannot select the words, you probably have a scanned PDF. It needs text recognition (OCR) before a reliable translation can be produced.
This guide explains how to translate a PDF to English, how to translate a scanned PDF, and how to keep the original document’s headings, tables, page order and meaning as intact as possible. You do not need to learn desktop publishing to read a foreign-language document. You do need to know which kind of PDF you have, which translation method suits it, and which details deserve a second look.
The important distinction is between reading a translated PDF and producing a translated PDF that can safely replace the original. For personal reading, a quick translation may be enough. For an application, employment contract, medical letter, academic paper or public-facing publication, the translation must be checked against the source; an attractive layout does not prove that the words are accurate.
Quick answer: translate a PDF on a computer
- Save a copy of the original PDF so you can return to it.
- Open the document and try to highlight one complete sentence.
- If the text is selectable, visit [Google Translate](https://translate.google.com/), choose Documents and upload the file.
- Set the original language or use language detection; choose English as the target.
- Download the translation and compare its pages with your source, especially numbers, tables, names and conditions.
- If the text is not selectable, perform OCR on a copy of the PDF, correct the recognized text and translate that text or the OCR-generated document.
- If page fidelity really matters, use the editable source file if one exists, translate that, review the text and export a new PDF.
The method is not quite the same for every document. A one-page school notice, a passport scan, a research report and a product manual create different risks. The rest of this guide shows how to make the right decision rather than forcing every PDF through one button.
What kind of PDF do you have?
A PDF is a presentation format. It records how a page should appear, but that does not tell you whether the visible letters are stored as real text. A born-digital PDF exported from Word will often contain searchable, selectable text. A PDF made by photographing or scanning a page can contain nothing more than an image. Some scans also contain a hidden OCR text layer, which may be incomplete or wrong.
Open the file and try three things. Select a word. Search for a distinctive phrase with the PDF reader’s Find command. Copy a short passage into a plain text editor. If the copied passage preserves the words in a sensible order, that is encouraging. If the cursor cannot highlight words, or copied text appears as random characters, you have discovered a text-extraction problem before attempting translation.
There are also mixed PDFs. A report may contain normal paragraphs, a scanned signature page, a chart with text baked into the image, and a photograph of a label. A translator might handle the selectable paragraphs while leaving words inside the images untranslated. Never assume that one successful translated page means the whole file was translated.
A useful habit is to label the file for yourself: selectable text, scanned image, mixed content, or editable source available. This small classification determines what to do next and saves time when a tool appears to fail.
Method 1: translate a selectable PDF with Google Translate
On a desktop or laptop browser, open Google Translate and choose the Documents tab. Select the source language if you know it, or use Detect language. Choose English for the output. Browse to the PDF, start translation, and download or view the completed result according to the options presented for your document.
Google’s published document-translation help currently describes support for PDF, DOCX, PPTX and XLSX files up to 10 MB, with PDFs limited to 300 pages. It also states that document translation is not offered on smaller screens or mobile through that workflow. Check the official [Google document translation instructions](https://support.google.com/translate/answer/2534559?hl=en) if the interface changes. These are product limits, not rules about how long a human translator can work on a document.
Once the translation is downloaded, resist the urge to delete the original. Put the two files side by side. Compare the first page, a middle page and the final page. Look at a few headings, every important table, and any page containing a warning or instruction. An automated tool can produce convincing English while making an error in a date, a condition or an implied subject.
For a long report, do a more systematic inspection. List the pages containing figures, amounts, negative values, footnotes and links. Open both versions at those pages. If the translation has changed the page count, compare sections by heading rather than assuming that source page 12 will always correspond to translated page 12.
Do not expect identical pagination. English may use more or fewer characters than the original language. Text can wrap differently, a word can become a longer phrase, and table cells can expand. A reasonable document translation can preserve structure while changing the exact line breaks. That is different from a broken file.
Method 2: translate a scanned PDF using OCR
A scanned PDF needs an extra step: optical character recognition. OCR turns the visible writing in a photograph or page image into text that software can search, select and translate. It does not understand the document perfectly. It guesses characters from image evidence, and those guesses must be checked.
Begin with the clearest source you have. If you can rescan an original paper page, keep it straight, evenly lit and sharp. Avoid cutting off the edges. A skewed page, blurred small print or a shadow through the middle can convert one word into another before translation even starts. If the scan is already available, check whether your PDF editor, scanner software or document-management tool offers OCR or Recognize Text. Use a trusted tool appropriate to the document’s confidentiality.
Save the OCR output as a new file, leaving the original untouched. Search for words that appear in the original. Copy text from a paragraph, a column and a table. Check that it reads in the right order. This is particularly important for two-column articles: OCR may accidentally alternate between columns or join unrelated rows. A translator receiving that distorted source has little chance of recovering the true argument.
Some OCR systems create searchable PDFs; others export a Word file or plain text. The editable export can be the better intermediate format when you need to correct errors before translating. You might have to repair headings, hyphenated line breaks and tables. This is ordinary document preparation, not unnecessary extra work.
Google’s document translation instructions explicitly warn that text inside images and scanned PDF pages may appear in the output without being translated. That is why an image-only document cannot be assumed to work in the same way as a text PDF. For a technical explanation of preventing recognition mistakes from becoming translation mistakes, read [Localize OCR and Document Scanning Without Changing What the Document Says](https://edukatesg.com/2026/09/23/translate-like-a-pro-localize-ocr-document-scanning-text-recognition-document-says/).
How to tell whether OCR has worked
Do not judge OCR only by how the page looks. A searchable PDF can visually resemble the original perfectly, yet its hidden text layer might contain incorrect words. Search for a word with an accent, a number and a phrase from the bottom of a page. Try copying a line with punctuation. Check names and uncommon technical terms; these are especially vulnerable to recognition errors.
Watch for lookalike characters. The digit 0 can be confused with the letter O; 1 with l or I; a decimal comma with a decimal point; a minus sign with a dash; Greek or Cyrillic characters with similar-looking Latin letters. These problems are not cosmetic when the file contains an account number, chemical symbol, equation, dose, reference identifier or measurement.
Check the order of sentences and paragraphs as well. A table read from left to right when it should have been read row by row can attach the wrong price to the wrong product. A footnote recognized in the middle of a sentence can change its apparent meaning. An OCR check must therefore be both character-level and structural.
If the OCR text is badly damaged, do not hide it behind a fluent translation. Correct the recognized source manually, obtain a clearer scan, or ask the document owner for an editable original. A translator cannot reliably preserve a meaning that the input no longer contains.
Method 3: translate the original Word document and export a fresh PDF
If the document began in Word, InDesign or another editing application, find out whether that source file still exists. Working from the source often gives you more control over style, paragraphs, tables, headings and fonts than trying to reconstruct them after PDF export.
For a DOCX file in Microsoft Word, a supported workflow is Review, Translate, Translate Document. Word then creates a translated copy in a separate document. You can review it before saving as a PDF. Microsoft describes this approach in its [Word translation help](https://support.microsoft.com/en-us/office/languages/translate-text-into-a-different-language). You can also translate selected text, but make sure a selective replacement does not create a mixed-language document by accident.
Review the translated Word file while it remains editable. Fix terms, inspect the table widths, check that page numbers and cross-references still make sense, and make sure text has not disappeared from text boxes or graphics. Then export the completed file as a new PDF and compare it with the original.
This route is especially useful for brochures, forms, training documents and manuals that will be reused. It gives you a chance to adjust the layout for English rather than squeezing English into a space designed for another language. The goal is to preserve readability and information, not to make every line break physically identical.
If you only have a PDF, converting it to Word may be useful but imperfect. Converted files can contain floating text boxes, fragmented paragraphs and awkward tables. Inspect the editable conversion before translation. You do not want a structural defect from the conversion stage to become an apparent translation problem later.
What does “without losing formatting” really mean?
People usually mean several different things when they ask to translate a PDF without losing formatting. Some want the same headings, bullets, tables and page order. Others want a print-ready document whose visual design closely matches the original. Those are not identical objectives.
Functional preservation means the reader can find the same sections, follow the same instructions, identify the same table headings and understand the same diagrams. Visual preservation is stricter: typography, spacing, page breaks, column widths and footnote positions should be close to the original. Even a correct translation can change the amount of space each sentence needs.
Begin by deciding which items are truly fixed. Legal page references, fillable-form fields, product codes, part numbers, branding positions, table relationships and page numbering may need special treatment. Ordinary paragraph line breaks are often flexible. A publishing-quality result may need layout work after translation, sometimes called multilingual desktop publishing.
For that specialised problem, the companion guide [Rebuild Translated PDFs and Print Layouts Without Breaking Typography or Reading Order](https://edukatesg.com/2026/09/18/translate-like-a-pro-multilingual-dtp-translated-pdf-print-layout-typography/) goes deeper. This article stays with the practical task of obtaining readable English from a PDF and verifying what you receive.
The first-page test
Before processing a 150-page report, test three pages: one normal text page, one difficult page with tables or figures, and one page near the end. Translate a sample using the route you intend to follow. Look for obvious failures, then decide whether the workflow should be used on the entire document.
This reveals problems that are otherwise expensive to discover later. Perhaps the main paragraphs translate well but charts do not. Perhaps the source is actually bilingual and language detection switches unpredictably. Perhaps text in a narrow table wraps into the next cell. Perhaps a disclaimer is missing from the output.
A small sample can also identify terminology that needs to stay consistent. If the document repeatedly refers to a named programme, regulation or component, decide how you will render that expression in English before translating dozens of occurrences.
You do not need a perfect sample. You need enough evidence that the method handles your document’s hardest ordinary features. A test on the easiest cover page is not a meaningful test of a complex report.
How to translate a PDF into English on a phone
On an iPhone or Android phone, the document-upload workflow in the Google Translate website may not be available. Do not keep hunting for a desktop-only button on a small screen. If your PDF contains only a few paragraphs that you need to understand, open it in a PDF reader, select a passage if possible, and send the text to your preferred translation app.
For a page you cannot select, you may be able to save or capture the visible page as an image and use an image-translation feature. Google Translate supports camera and gallery image translation on phones; its [image-translation instructions](https://support.google.com/translate/answer/6142483?co=GENIE.Platform%3DiOS&hl=en) explain the mobile workflow. Check the complete sentence, not only isolated words, because small screenshots can omit the surrounding context.
For an entire report, a computer is normally more practical. You can compare pages side by side, export editable OCR results, correct errors and save a final document. On a phone, repeating the screenshot method for 80 pages is slow and creates many opportunities to lose page order.
Never confuse a translated screenshot with a complete translated document. It may help you read one page immediately, but it is not a dependable replacement for the entire PDF. Keep the original accessible while you work.
What if the PDF has two columns?
Research papers, newsletters and brochures often use multiple text columns. Their visual reading order may be obvious to a person but unclear to extraction software. Copy a paragraph from the top of the left column and see whether the next copied sentence truly follows it. If the result jumps to the right column and back again, your extraction is unsuitable for direct translation.
Where possible, export with reading-order-aware OCR or convert the layout into a single logical text flow while preserving section labels. You can translate the clean text first, then reintroduce columns in the final layout. That is often more reliable than asking one tool to solve OCR, translation and complex graphic placement simultaneously.
Figures introduce another complication. A diagram label can be spatially meaningful. “Input” on the left and “Output” on the right is not just two floating words. If a translation tool extracts them without their positions, keep a screenshot of the original figure and annotate which translated label belongs to which part.
For accessibility, the reading order is important too. A PDF that looks correct can still be confusing to a screen reader if the text sequence is broken. When preparing a shareable final PDF, inspect both the visible layout and the underlying text order.
How to preserve tables and numbers
Numbers should be checked separately from ordinary prose. A language translation may not alter the numerical value, but it can change punctuation conventions or reflow a table. A decimal comma in one locale may appear as a decimal point in another. A date written 03/04/2026 is ambiguous unless you know whether it means 3 April or 4 March.
Read each column heading as a unit with its data. If a price column appears under the wrong product after reflow, the document is functionally wrong even if every price was copied faithfully. Look for missing rows, duplicated rows and wrapped labels that appear attached to another cell.
Pay attention to units. A figure of 5 mg must not silently become 5 g. A rate per month must not be described as a rate per year. In a statement such as “no more than 25 units,” changing the limit to “at least 25” reverses the instruction. These are translation integrity checks, not optional proofreading niceties.
For invoices and budgets, reconcile totals where possible. Check that negative amounts stay negative, that the currency label remains attached to the amount, and that a translated “net” or “gross” term matches the document’s calculation. When in doubt, compare the original table with the output cell by cell.
Names, addresses and official identifiers
Do not automatically translate everything that looks like language. Personal names, business names, building names, programme titles and reference numbers may need to be preserved exactly or given an official English form. A machine tool may invent a plausible-looking English name that is not the person’s registered name.
If an official institution already publishes its English name, use that verified form rather than translating each word independently. If no authoritative form exists, keep the original name alongside a careful explanatory rendering when that helps the reader. Do not change passport numbers, application references, registration codes or QR destinations.
Addresses have practical functions. Translating a street name literally might make an address impossible to use for deliveries or appointments. Preserve the postal identifier and match the form used by the relevant local authority. A helpful English explanation can sit beside the original address without replacing it.
It is worth creating a short document glossary before final review. Record the original term, approved English equivalent, and whether the item must remain unchanged. This is far more effective than debating the same name every time it appears.
Footnotes, citations and links
A translated PDF may still be used for research or professional reading. That means footnote markers, document references, bibliography entries, tables and links need to survive. Check that footnote number 4 still leads to the fourth note, not the fifth, and that a source title is not translated in a way that conceals the original publication.
A citation may include a non-English article title and an English translation in square brackets. The exact convention depends on the relevant style guide. The essential point is to keep the work identifiable. Readers should be able to find the original source, even if the translated title helps them understand its subject.
Hyperlinks deserve manual inspection. If the translated file contains a button labelled “Download form,” ensure it still opens the intended form. A visually correct label with a broken or changed destination can make the translated document unusable.
When the output is for internal reading, not every citation needs editorial reconstruction. When the output is destined for public distribution, each reference is part of the reader’s ability to verify claims and should be checked accordingly.
What to do with charts, diagrams and image labels
Charts can contain embedded text that ordinary document translation skips. Look carefully at axis labels, legends, chart titles, units, data source captions and notes beneath the image. Leaving one axis in the source language may be acceptable for your personal study, but a public-facing report should not pretend that the graphic has been fully translated.
When necessary, create a translated text key outside the chart rather than trying to paint English over a complex image. For example, identify each source-language legend item and its English equivalent in an adjacent list. This is slower than a one-click output but safer than placing a label in the wrong position.
Do not change the data while relabelling. A chart labelled “Monthly output” cannot be renamed “Annual output” just because an automatic translator produces a more familiar expression. Preserve the underlying measurement, timescale and comparison.
For critical diagrams, have a domain-aware reviewer inspect the correspondence between captions and visual features. Language correctness and diagram correctness are inseparable when an instruction depends on an arrow, a warning symbol or the location of a component.
A worked example: a school notice
Imagine a two-page notice in another language about a school excursion. It includes the location, a meeting time, payment instructions, a consent deadline and a contact number. You translate it and receive smooth English. That is a useful start, but there are several distinct facts to check.
First identify the action required. Must a parent sign a form, reply by email, pay a fee, or do all three? Then inspect the dates. If the notice mentions both the deadline and the excursion date, a translation that reverses them is not a minor grammatical issue. Check whether a condition such as “only if the child is attending” has been preserved.
Now look at the practical fields. The phone number and bank reference should normally remain intact. A place name may have an official English form or may need to be preserved as originally written. If the trip requires special clothing, a mistranslated negative instruction could produce the opposite result.
Finally ask a reader who has not seen the original to explain what they must do. If they cannot identify the deadline, meeting point or required permission, the translation has not completed its practical job, even if the sentences appear elegant.
A worked example: a product manual
A manual may include numbered steps, safety warnings, part names and diagrams. Suppose the original instruction says to disconnect power before removing a cover. If the English version suggests removing the cover first and disconnecting power afterward, the order has changed. That is a substantive error.
Translate imperative verbs and conditions carefully. “May” can indicate permission or possibility. “Must” is not the same as “should.” “Before” and “after” must remain anchored to the correct actions. Where a tool produces awkward language, the fix must make the instruction clearer without changing its sequence.
Compare part numbers with the source. A named part can be translated descriptively, but the manufacturer identifier must remain stable. If the manual contains a diagram, cross-check each part label rather than assuming the surrounding paragraph is enough.
For equipment involving significant safety risks, use qualified technical translation and review. Automated output may support initial comprehension but is not a suitable final authority for hazardous procedures.
A worked example: an academic paper
An academic PDF may contain an abstract, methods, results, discussion and references. Read the translated abstract for orientation, but do not treat it as a substitute for the results tables or methods. Words such as “significant,” “associated,” “caused” and “estimated” can carry technical meanings that should not be flattened into everyday English.
Compare the claim to the evidence. If the source says an outcome was associated with a variable, a translation that states the variable caused the outcome is stronger than the evidence permits. Check whether confidence intervals, uncertainty and limitations remain visible.
Maintain author names, publication details and identifiers. If OCR has damaged Greek letters or formula symbols, correct them before reasoning about the reported result. A numerical subscript can be important, and a dropped minus sign can reverse a value.
For serious research use, seek a bilingual or subject-qualified reviewer. A quick PDF translation is useful for screening whether a paper is relevant; it is not automatically sufficient for quoting or republishing its findings.
Common problems and how to fix them
The translated PDF is blank or most pages remain in the original language
Check whether the source contains a real text layer. The file may be an image-only scan or have text inside drawings. Try selecting or searching a sentence. If that fails, perform OCR and inspect the result before translating. If only chart labels remain untranslated, treat those graphics separately.
Characters become squares or strange symbols
This is usually an extraction or font issue rather than a language issue alone. Check whether the source text can be copied correctly. If copied text is corrupted, a translator will receive corrupted input. Use a better export, another OCR method or the editable original; then verify the fonts used in the final English document.
Paragraphs appear in the wrong order
The PDF may have columns or floating text. Try exporting to an editable format and correct the reading order, or use tools that explicitly preserve document structure. Translate after the text order is repaired. Do not rely on visual appearance as proof of underlying reading order.
The translation sounds fluent but the meaning seems odd
Return to the complete source sentence and its paragraph. A single word may have multiple meanings, and automated translation can choose the wrong one. Check technical terms, pronouns, negation and what each condition modifies. Use a reliable dictionary or subject reference where needed, then correct the passage.
Table text is cut off
Reflow the table in an editable file, shorten only wording that can be shortened without loss, or change the layout. Do not shrink text until it becomes unreadable merely to preserve the original page count. For a print-ready document, the solution may require manual desktop publishing.
The service rejects the upload
Check size and format. Google Translate’s documented website workflow gives explicit file size and page limits; different products have different restrictions. You may need to use the editable source, split a document into responsibly labelled sections, or choose an appropriate document translation service. Avoid splitting across pages in a way that separates headings from their explanations.
How to check a PDF translation without knowing the source language well
You can still test structural integrity even if you cannot fully evaluate linguistic accuracy. Count the pages and sections, inspect the table rows, compare numbers and names, and verify that dates and identifiers did not disappear. Mark source passages that are important to your decision and ask a qualified speaker to review those passages.
Do not mistake back-translation for a definitive accuracy test. Translating English back into the source language can reveal suspicious wording but can also produce different yet equivalent phrasing. If the English version says “Payment required before registration” and the original might say “Registration required before payment,” investigate the source directly.
For high-stakes work, identify what you need from the translation. Perhaps you only need to know where to submit a document. Perhaps you need legally usable English. These are different quality requirements. If a certified or officially accepted translation is required, check the recipient authority’s rules before paying for or submitting a machine-generated result.
A practical review record can be short: source filename, translation date, tool or translator, language pair, pages checked, unresolved passages and reviewer. That small trail helps if you later discover an error.
Privacy and permission before uploading
A PDF might contain personal data, financial records, student information, confidential business plans or protected research. An online translation service is not a private extension of your computer. Uploading may be governed by the service’s current terms, product configuration and retention policies.
Before using any external tool, confirm that you are allowed to send the document to it. For confidential or regulated documents, use an approved organisational workflow or qualified service with suitable safeguards. Do not paste account credentials, medical records, passport scans or unpublished proprietary information into an arbitrary website because it promises free translation.
Where you can, remove unnecessary personal information from a working copy without altering the facts you must translate. If redaction would make a document unusable, preserve a secure original and seek an approved workflow rather than risking exposure.
Privacy and translation quality are separate questions. A tool can be linguistically useful but unsuitable for a sensitive file. Choosing a safe process is part of doing the job correctly.
How long does a PDF translation take?
A one-page selectable text PDF may translate very quickly in a suitable tool. A hundred-page scanned manual may require substantial OCR correction, terminology research and layout reconstruction. The difference is not merely the word count. It is the amount of structure and verification required.
For a personal reading copy, concentrate on the sections you actually need. Translate the executive summary or first relevant chapter first. For a publishable file, budget time to compare the output with the original and conduct a target-language read-through. Translation is finished when the document can be used, not just when the download button appears.
It may help to separate three deliverables: a rough reading translation, a reviewed translation, and a production-ready PDF. They have different purposes and need not cost the same amount of work. A clear scope prevents disappointment.
If someone offers perfect one-click formatting and accuracy for every document, be sceptical. PDFs vary too much in text encoding, scans, layout and subject matter for an unconditional promise to be credible.
A reusable PDF translation checklist
- Confirm whether the PDF has selectable text.
- Identify the original language and whether multiple languages appear.
- Decide whether you need a reading copy or a publication-ready copy.
- Save the source and work on a copy.
- Apply OCR first when the file is scanned or text extraction is defective.
- Check the OCR reading order before translation.
- Keep important names, identifiers, numbers and units traceable.
- Review tables, figures, headers, footnotes and links.
- Correct terms consistently across repeated sections.
- Check conditions, dates, deadlines and instructions.
- Inspect fonts, page breaks and text overflow.
- Verify permissions and data-handling requirements.
- Save the final output with a clear filename and version.
- Keep a short record of unresolved questions or review decisions.
These checks are more useful than asking whether the PDF looks “professional.” A document can look beautiful while containing a wrong number. A rough-looking translated copy can still be useful for understanding, provided you know its limits and retain access to the original.
Frequently asked questions
Can I translate a PDF into English for free?
You can often use no-cost translation features for ordinary files within their supported limits. Some services have file, page or size restrictions, and scanned documents may need OCR. Check the current product instructions. For sensitive or official work, the relevant question is whether the result is appropriate and accepted, not just whether the upload is free.
Can Google Translate translate a scanned PDF?
Its document-translation help warns that text in images and scanned PDF pages is not translated by that workflow. Create and check a usable text layer with OCR, or use an image translation workflow for selected pages. Verify recognition errors before relying on the English output.
Will the translated PDF look exactly like the original?
Not always. Language expansion, fonts, tables and line breaks can change. You may preserve headings and page order yet need manual formatting for a faithful print layout. If exact design matters, translate the editable source and perform layout quality assurance.
Can I translate a PDF on my iPhone?
You can often translate selected text or use image translation for screenshots of pages. The desktop-style Google Translate document upload workflow is not generally supported on mobile screens. For long documents, a computer and an editable intermediate file are more reliable.
How do I translate a password-protected PDF?
Only do so if you are authorised to access and process it. Ask the owner for a permitted copy or an editable source. Do not use questionable password-removal services; they may expose the document and can breach the owner’s restrictions.
How do I translate a PDF while keeping hyperlinks?
Check the output links individually. Some conversion routes preserve clickable links, while others convert them into plain text or change their targets. Preserve meaningful link labels and destinations, then test the final exported file.
Is an automatically translated PDF legally valid?
That depends on the purpose, jurisdiction and receiving organisation. A reader’s convenience translation is not the same thing as a certified or officially accepted translation. Ask the relevant authority what qualifications, certification, notarisation or document format it requires.
What is the best way to translate a PDF with many tables?
Use an editable source when possible. If only the PDF exists, extract text with attention to rows, columns and reading order; translate; then rebuild and inspect each table. Do not assume that a visually plausible table has preserved the original data relationships.
Where to continue
If you need the bigger principles behind accuracy, context and verification, read [Master Art of Translation](https://edukatesg.com/2026/09/18/master-art-of-translation-complete-system-moving-meaning-between-languages/). If a translated word feels plausible but still wrong, the [Vocabulary Learning Hub](https://edukatesg.com/vocabulary/) helps you examine meaning and usage more closely. The practical sequence for a PDF remains simple: identify the file type, obtain trustworthy text, translate, then check the facts and layout before relying on the result.