Compress PDF

That depends on the mode, and the difference is worth understanding before you pick one. Visual renders each page to a bitmap at a reduced resolution and stores that picture — everything on the page becomes part of the image, so text stops being selectable and will soften if you zoom in. It is the right choice for scans and photo-heavy documents, where it can cut the size dramatically. Native copies every page exactly as it is, so text and vector artwork stay perfectly sharp and selectable; its savings come only from rewriting the file structure, so they are modest and sometimes nil. Individual embedded images are not resampled in either mode.

Less than you might expect — typically a few per cent rather than tens. The page is stored as a JPEG, and JPEG already compresses colour information far more aggressively than brightness, so removing the colour does not free up as much as it sounds like it should. It is worth switching on when the original is genuinely black and white, and it is not the lever to reach for when you need a big reduction. Resolution is. Grayscale also only changes pages that are re-rendered. It has no effect on pages kept as they are.

Switch on Target File Size, set 2 MB, and keep the mode on Visual. It is a measured search, not an estimate: the tool renders the document, weighs the actual bytes, and adjusts quality and resolution up to five times to land just under your limit while giving away as little quality as it can. If even the most aggressive setting stays over budget it tells you and hands back the smallest result rather than pretending it succeeded.

Drop the PDF in and start with the Balanced preset, which re-renders the pages in Visual mode and gives the biggest reduction on scans and photo-heavy files. If the PDF is mostly typed text, try the Native preset instead: it keeps the text selectable and sharp, although the saving is smaller. For email, Gmail caps attachments at 25 MB and many work mail systems are stricter, so Target File Size set to 10 MB clears most inboxes. When a file is still too large, a shared link beats a smaller and blurrier attachment.

Switch on Target File Size, enter the limit, and keep the mode on Visual, because only Visual can trade image quality for bytes; Native copies pages unchanged and has nothing to trade. The smaller the target, the more pages matter: a few scanned pages fit in 200 KB or even 100 KB with readable text, while a 40-page scan at 100 KB turns to mush. In Visual mode the text becomes part of an image and can no longer be selected, so if the portal needs searchable text, check the result before you submit it.

Some of it, and not because of a setting. The output is built as a brand-new document with your pages copied into it, so these document-level parts do not come along: JavaScript, attachments, bookmarks, the XMP metadata, incremental-save leftovers and the accessibility tag tree. That last one is worth flagging if the file needs to stay screen-reader friendly. What is stored on a page stays with pages kept as they are in Native mode: embedded page thumbnails, app data saved on the page, link and page actions, file-attachment annotations, and form fields, which stay fillable. The Remove Document Metadata toggle controls one specific thing on top of that: whether the title, author, subject and keywords are carried over or dropped.

Because rendering it would have made it bigger. A text or vector PDF stores a few kilobytes of glyph instructions per page; turning that page into a photograph can easily cost more. When that happens the tool tries again without re-rendering, and if that still does not beat the original it hands your file back untouched under its own name rather than shipping something larger and calling it a saving. When the pages will be re-rendered as JPEG, the rough guess beside each file warns you when this is likely.