What “Print to PDF” actually does

Two different routes to the same file extension

Most applications offer both, often within two menu items of each other, and they are not the same operation.

Export, or Save as PDF. The application writes the PDF itself, or hands a writer everything it knows: this run of text is a heading, this word is a link to that address, this is the table of contents, these are the form fields. All of it has somewhere to go in a PDF, because the format has a place for each.

Print to PDF. The application talks to a printer. It says: put these glyphs at these coordinates, fill this rectangle, draw this image here. A virtual printer collects that description and writes a PDF of it. Nothing else is in the conversation — not because the writer discarded it, but because the print interface was designed to drive a machine that puts ink on paper, and a machine that puts ink on paper has no use for a bookmark.

Both Windows’ Microsoft Print to PDF and the Save as PDF item inside the macOS print dialog go down the print route. A browser’s own Save as PDF destination is a third thing again: the browser renders the page and writes the PDF directly rather than going through the operating system’s print system, which is why it usually keeps hyperlinks when a print driver does not. Behaviour varies by application, driver and version, so treat everything below as what to check rather than what to assume.

What does not survive the trip

“Usually” is doing real work in that table and it is not hedging for its own sake: drivers differ, applications differ, and the only reliable statement is about the mechanism rather than about any particular pair of them.

The loss with consequences beyond convenience is the tags. A structure tree is what tells assistive software that this line is a heading, that this is a list, and in what order the page should be read.

A document printed to PDF has no structure tree at all, which means it cannot meet an accessibility requirement no matter how good it looks — and the failure is invisible to everyone who is not using a screen reader. If the document is going anywhere that has an accessibility obligation, the print route is not available to you. What makes a PDF accessible covers what a tagged file needs.

The second loss worth planning for is the links. Nothing on this site — and, in general, nothing short of an editor — recreates link annotations from text that merely looks like a URL. If the links matter, go back and export.

Why the result is not quite the size you designed

Two separate things about the printer get baked into the file.

The paper size is the driver’s, not the document’s. An A4 document printed on a machine whose default is Letter comes out Letter, with the content scaled to fit. This is the ordinary A4-against-Letter problem arriving through a side door, and PDF page size and orientation covers the shape of it.

The imageable area is smaller than the sheet. A physical printer has to grip the paper, so there is always a margin it cannot reach, and the driver declares a printable rectangle inset from the page. Many applications default to shrinking the content to fit that rectangle — “fit to printable area”, “shrink oversized pages”, “scale to fit”. A PDF has no physical edge and no reason to have that margin, but the scaling was applied before the PDF writer ever saw the page, so it is in the output permanently. The symptom is a page a few percent small, sitting inside a white border, with every measurement slightly wrong.

To avoid it: set the paper size explicitly in the print dialog, and set the scale to 100% or “actual size” rather than any fit option. Then check with page count and file info, which reports the page dimensions of what you actually produced. If you already have a file that came out wrong, resize PDF can scale it back onto the right page, though it is scaling a scaled page and the text will not be quite as crisp as re-exporting.

When printing is the right choice anyway

It has genuine uses, and it is worth being fair about them.

• The application has no export. Plenty do not. Printing is then the only route, and it works.
• You want what the printed page would look like. Print layout adds headers, footers, page numbers and page breaks that an export does not, and sometimes that is exactly the artefact you need — a record of the thing as it would have been printed.
• You want the form flattened. Printing a filled form makes the values part of the page. It works, but flatten PDF does the same job deliberately and keeps everything else about the file intact.

And one habit worth breaking. Printing a PDF back to PDF, to “clean it up” or to get past a fussy system, is reached for far too early. It is a complete re-render: it can grow the file, it strips the structure the document had, and if the driver decides to rasterise a page — which some do when they meet transparency they cannot express — the text stops being text and you find out later. It will also drop a permissions password, which is a question about what you are entitled to do rather than a feature.

Three checks on the result

1. Is the text still text? Run the output through PDF to text. If nothing comes back, the driver rasterised the page and you now have pictures. That is recoverable only by going back to the source, or by OCR, which is a new set of guesses rather than the original text.
2. Is the page the size you asked for? Page count and file info reports the dimensions. A page 2 to 4% under the nominal size is the shrink-to-fit signature.
3. Do the links work? Click one. If they are dead, no post-processing here will revive them — export from the source instead.

If the file has grown unreasonably, compress PDF will usually bring it back down, because a print driver that embeds full fonts or re-encodes images generously leaves a good deal of slack. How PDF compression works covers which mode to reach for.