A file saved by Pages '09 or earlier holds QuickLook/Preview.pdf, the whole document rendered by Apple, and this tool takes it out untouched. That is an extraction rather than a conversion, which is why the result is exact. Pages 5 and later, the versions since 2013, store only a preview image of the first page, so a current file has no PDF to extract and the tool says so.
For a current file, Pages for iCloud opens it in any browser with an Apple Account and downloads it as PDF. The other route is to ask the sender for an exported PDF.
What is actually in a .pages file
It is not a document in the way a .docx is. It is a zip of an Apple package, and since iWork '13 the document body is stored as IWA: Apple's own protocol buffers, framed with Snappy compression, with an internal schema that is undocumented and changes between versions. Nothing outside Apple's software reads it reliably, and tools that claim to are guessing.
What sits beside that body depends on the version. Pages '09 and earlier stored the document as index.xml and, with previews enabled, also wrote a full PDF render of it into the package, so that Finder, Quick Look and Mail could show it without launching the app. Pages 5 and later write a preview.jpg of the first page instead, which is enough for a thumbnail and not for reading the document. The two never sit in the same file.
Index/*.iwa
The real document in a current file, in Apple's private binary format. Effectively unreadable elsewhere.
QuickLook/Preview.pdf
Only in files from Pages '09 and earlier: the whole document as a PDF, rendered by Apple with the right fonts. This is what the tool extracts.
Data/
Images and media placed in the document, in their original form.
Metadata/
Which version of Pages wrote the file, and the identifiers it uses for syncing.
preview.jpg
What current Pages writes where the PDF used to be: a picture of the first page. The tool offers it as a one-page PDF, labelled as a picture of page one, beside the steps for a full export.
Why this beats converting
Every other route has to rebuild the document. It reads whatever structure it can recover, guesses at the fonts (which are usually Apple's and not installed on your machine), re-flows the text, and hands you something that resembles the original. Tables shift, line breaks move, and a carefully placed image ends up a page later.
Where the PDF exists, extraction has none of those failure modes because there is no layout decision to make. The PDF was produced by Pages itself, on the author's machine, with the author's fonts. What you download is byte for byte what Apple wrote. This tool does not even re-save it through a PDF library, because re-saving is a chance to change something and there is nothing to gain by taking it.
Parse and rebuild
What most converters attempt. Fonts substituted, spacing drifts, tables break, and complex layouts come back wrong in ways that are hard to spot until someone points at page four.
Extract what Apple rendered
Exact fonts, exact spacing, exact pagination, because it is the same PDF you would get by printing from Pages on the author's Mac. Nothing is inferred.
When the preview is not there
That is every file saved by Pages 5 or later, which means every file a current Mac, iPhone or iPad produces: the package holds a preview.jpg and no PDF. An older file can lack the PDF too, if it was saved with Include preview in document switched off.
Nothing outside Apple's software can rebuild the document from such a file, and any site that promises otherwise is either guessing at the layout or quietly failing. This tool does not turn the preview image into a PDF. It tells you there is nothing to extract rather than producing something broken.
Two routes work. Pages for iCloud opens any .pages file in a browser, and its Download a Copy command saves a PDF; that means uploading the document to Apple, which matters for a confidential file. Or whoever has Pages chooses File, then Export To, then PDF, and the result is better than any conversion would have been anyway.
A .key presentation and a .numbers spreadsheet use the same package layout and have the same history: files from the '09 generation can carry a PDF, and current ones carry an image. The tool accepts both, with the same result. They each have their own surprises, though, so there is a page for each: Keynote to PDF covers the slide shape and the notes the render leaves out, and Numbers to PDF covers why the page breaks were the sender's decision rather than yours.
Questions people actually ask
What people ask when a .pages file lands on a Windows machine.
01Can I edit the PDF it gives me?
It is an ordinary PDF, so you can annotate, sign, fill fields and reorder pages with the tools on this site. You cannot get back to an editable word processor document, because the PDF records the finished layout rather than the structure that produced it.
02Is the text selectable?
Yes. Apple's preview is real text with embedded fonts, not a picture of the page, so search, copy and select all work.
03The page count is lower than the document I was told about.
The preview reflects the document as it was last saved. If the sender edited after that without saving, or saved from a different device, the preview can lag. When the count looks wrong, ask for an exported PDF.
04What about .pages files from iCloud?
They are the same package as any current file, so they hold a preview image and no PDF, and the tool has nothing to extract. Open the document in Pages for iCloud in a browser instead and download it as PDF from there, which needs no Mac.
05Does the file get uploaded?
No. The zip is opened in your browser with JavaScript and the PDF inside is handed to you directly. Nothing is transmitted.