The PDF is read and translated on your device by the translator built into your browser. Nothing is uploaded.
How the layout is kept
Each page is drawn as a picture, and the original text is painted over with the colour behind it. The text is grouped into paragraphs and columns, each paragraph is translated as a whole so the sentences read naturally, and the result is spread back over the original lines. Where a translation runs longer, the type is made smaller so it fits the original width. Headings, tables of figures, images and page design stay where they were.
Which browsers can translate?
Desktop Chrome 138 or newer has a built-in translator. The first time you use a language pair it downloads a small language pack, and after that it works offline. Other browsers do not have one yet; you can still save the PDF’s text and translate it elsewhere.
What you get
A PDF that looks like the original, with the translation in place. For languages written in Latin, Greek or Cyrillic letters the translated text is real, selectable text. For other scripts, such as Japanese, Chinese, Arabic or Hindi, it is drawn into the page image so it shows correctly, and the reflowed document carries the selectable text. The reflowed document is a simple web page with headings and paragraphs that you can read on any screen, copy from, or print to PDF.
Limits
Scanned PDFs have no text to translate; make them searchable first. Text inside images, charts and form fields is left as it is. Machine translation can get names, terms and tone wrong, so check anything important.