Updated September 2026
"Without losing quality" means different things depending on what's actually inside your PDF — a page of typed text and a page that's a scanned photograph respond to compression very differently. Here's what changes, what doesn't, and how to know which applies to your specific file.
Recompressing embedded images with tighter (but still lossless-to-visually-identical) encoding, recompressing fonts, and removing duplicated internal objects doesn't touch how a page looks or reads. Text renders exactly as sharp as before, at any zoom level, and stays fully selectable and searchable. If your PDF is mostly typed text — a contract, a report, an invoice — this pass alone usually accounts for most or all of the size reduction you'll see, and there's no visible trade-off at all.
For image-heavy PDFs, a second technique re-renders each page as a flattened image at a set resolution (150 DPI here) and JPEG quality, then compares the result to the structural pass. This is genuinely lossy — the page becomes a picture of itself rather than its original content — and it's only used when it actually produces a meaningfully smaller file. At 150 DPI, text and line art in the rendered image are sharp enough for normal screen reading and standard printing, but if you zoom in far past 100%, you will see JPEG artifacts that weren't there in a vector original.
Whichever pass runs, the result only replaces your original if it's actually smaller — and the lossy pass specifically only wins if it beats the lossless pass by a meaningful margin, not just barely. If a file is already well-optimized and neither pass improves on it, you get your unmodified original back rather than a needlessly re-rendered, lower-quality copy just for the sake of "compressing" it.
After compressing, open the downloaded PDF and try two things: zoom in on a text-heavy page and try to select a sentence. If the text highlights normally, that page went through the lossless pass. If you can't select anything and the "text" is really part of a flattened image, that page went through the visual re-render — expected and usually harmless for a scan, worth noticing for anything you might need to search or copy from later.
For most everyday use — emailing, uploading, archiving a copy — the automatic result is the right trade-off: smaller file, no visible difference. For documents where exact fidelity genuinely matters (a legal exhibit with fine print, a technical drawing with fine detail, an image you'll need to zoom into later), check the compressed output before relying on it, the same way you'd sanity-check any automated tool. Full mechanics of both passes, including the exact thresholds used, are on the How Compression Works page.
The lossy pass isn't triggered by "does this file have images" alone — it's triggered by whether re-rendering actually beats the lossless result by a real margin (at least 8% smaller here, not just marginally smaller). That margin exists specifically to avoid trading meaningful quality for a trivial size saving. A file that's borderline between the two techniques is more likely to keep its lossless, full-fidelity result than to be pushed into a re-render for a small gain — the tool is deliberately biased toward keeping quality when the size difference wouldn't be worth it.
If your PDF is mostly words, expect full quality preservation as the default outcome, not the exception. If your PDF is mostly scanned or photographic pages, expect the visual re-render to run — and expect that to be the right trade-off, since those pages usually have far more resolution than needed for normal reading in the first place. See why scanned PDFs shrink the most for that case specifically.
Can I choose to skip the lossy pass entirely? Not currently — the tool automatically picks whichever result is smaller. If preserving exact fidelity matters more than file size for a specific document, keep the original alongside the compressed copy.
Will compressing the same file twice make it smaller each time? No — running an already-compressed PDF back through the tool typically returns it unchanged, since there's nothing further to gain; you won't see it shrink repeatedly.