You exported a 47 MB PDF from your design tool and your client’s inbox rejects anything over 25 MB. Or you’re trying to upload a scanned contract to a portal that caps at 10 MB. You search “compress PDF online free” and the first ten results all claim the same thing: shrink your file by 90%, no quality loss, in seconds.
Some of that is true. Most of it depends on what’s actually inside your file.
This guide covers what PDF compression actually does under the hood, what kind of size reduction is realistic for the document you have in front of you, and how five popular tools (iLovePDF, Smallpdf, Adobe Acrobat online, PDF24, and imisspdf) compare in 2026 — including the architectural difference that matters when the document is sensitive.
What PDF compression actually does
A PDF is a container holding three kinds of payload: vector content (text, line art, fonts), raster content (embedded images), and stream metadata (the objects that describe how everything is laid out). Each of these compresses differently.
When a compressor processes your PDF, it does some combination of the following:
1. Stream re-encoding. PDFs store content in “streams” that may or may not already be compressed. A poorly exported PDF can contain uncompressed streams or streams using suboptimal filters. The compressor re-encodes them with the best available filter (typically Flate, which is the same deflate algorithm zip uses). This is lossless. Size win: usually 5-15% on office documents, 0-5% on already-optimized PDFs.
2. Object deduplication. A PDF that was assembled from multiple sources often contains duplicate font subsets, duplicate image resources, or duplicate XObject definitions. The compressor merges these into single shared objects. This is lossless. Size win: 5-20% on merged or assembled documents, near zero on cleanly-exported ones.
3. Font subsetting. A PDF that embeds full fonts contains every glyph in the font, even ones the document never uses. Subsetting keeps only the glyphs actually referenced. This is lossless. Size win: 10-30% on documents that embed full TTFs, near zero on documents that already subset.
4. Image resampling. This is where the big wins live, and also where compression becomes lossy. A 600 DPI scan rendered at A4 page size contains roughly 35 million pixels per page; at 150 DPI that drops to about 2.2 million pixels per page — a 16x reduction in raw image data before any encoding. The compressor downsamples each embedded image to a target DPI, typically 72, 96, 150, or 300. This is irreversible for the resampled image but invisible at screen resolution if the target DPI matches the viewing context.
5. Image re-encoding. After resampling, each image is re-encoded — usually with DCT (the JPEG algorithm) at a configurable quality level. Going from quality 95 to quality 75 typically halves the image data with no visible change at screen sizes. Going to quality 50 produces visible block artifacts in flat areas but is acceptable for thumbnails. JPEG2000 (JPX) offers better quality-per-byte than JPEG but is supported by fewer tools and viewers; per the Ghostscript documentation, even mainstream PDF engines don’t fully support JPEG2000 encoding.
6. Metadata pruning. PDFs accumulate metadata over their lifetime: edit history, XMP packets, embedded thumbnails, attached files, JavaScript actions, form-field state. Aggressive compressors strip what’s not strictly required to render the document. Lossless from a rendering perspective; lossy if you care about the metadata.
When a tool advertises “compress your PDF by 90%”, they’re almost always relying on aggressive image resampling and re-encoding. That’s a real win — but it only applies to PDFs that contain large images. Re-running the same compressor on a text-only PDF will give you 0-5%.
Realistic size reductions by document type
Here’s what to actually expect, based on industry-published numbers and what we see in our own usage telemetry:
| Document type | Lossless only | Lossy (balanced) | Lossy (aggressive) |
|---|---|---|---|
| Text-heavy report (Word export) | 0-5% | 5-10% | 10-15% |
| Office document with charts | 5-15% | 20-30% | 30-40% |
| Mixed text + photos | 10-20% | 40-60% | 60-75% |
| Scanned document (300 DPI) | 5-10% | 50-70% | 70-85% |
| Scanned document (600 DPI) | 5-10% | 70-85% | 85-92% |
| Photo album / portfolio | 5-10% | 60-75% | 75-90% |
| Already-compressed file | 0-2% | 0-5% | 0-15% (with quality loss) |
The headline “up to 90% reduction” claims you see on tool homepages are real, but they describe the best case (a high-DPI scan resampled aggressively), not the typical case. Smallpdf’s own documentation is honest about this: image-heavy PDFs can shrink by up to 90%, while text-only documents typically see 20-40% — and that 20-40% number assumes there was meaningful redundancy to remove in the first place.
When compression helps — and when it doesn’t
Compression helps a lot when:
- The file came from a scanner (any DPI, any color depth)
- The file embeds high-resolution photos or screenshots
- The file came from a presentation tool that embeds source images at full resolution
- The file went through several rounds of editing in different tools (lots of accumulated cruft)
- The file was assembled from multiple PDFs that share fonts
Compression doesn’t help much when:
- The file is a clean export from a modern word processor (already small)
- The file is text-only with no embedded images
- The file already went through an aggressive compression pass (re-compressing yields little, can degrade quality)
- The file is dominated by a small number of vector graphics (already efficient)
If you compress a 200 KB Word export and the result is 195 KB, that’s not a tool failure — it’s a sign the file was already close to its minimum. The fix in that case isn’t a different compressor, it’s accepting that the file is as small as it usefully gets.
The five popular tools, compared
We compared the five most-clicked “compress PDF” results on a realistic mixed document (a 12-page PDF combining text, photos, and a couple of full-page screenshots, starting at 18.4 MB).
| Tool | Result size | Reduction | Architecture | Free tier limits | Watermark |
|---|---|---|---|---|---|
| iLovePDF | 5.2 MB | 72% | Server upload (Spain) | 25 MB max, ads | None |
| Smallpdf | 4.8 MB | 74% | Server upload (Switzerland) | 2 docs/day, 5 GB max | None |
| Adobe Acrobat online | 5.6 MB | 70% | Server upload (US/EU) | Sign-in required | None |
| PDF24 (online) | 5.4 MB | 71% | Server upload (Germany) | 100 MB max | None |
| imisspdf | 5.0 MB | 73% | In-browser (no upload) | Device RAM only | None |
The compression ratios are remarkably similar because all five tools use variants of the same underlying techniques: image resampling to ~150 DPI, JPEG re-encoding at quality ~75, font subsetting, and stream re-encoding. There’s no proprietary magic in the compression itself.
Where the tools actually differ is everything around the compression — the upload step, the retention policy, the file size cap, the pricing structure, and whether the file leaves your device at all.
iLovePDF
The most-trafficked PDF tool on the internet (~287M monthly visits per Similarweb). Spanish company, ISO/IEC 27001 certified, GDPR-compliant, files deleted within 2 hours per their security page. The 25 MB cap on the free tier is the practical friction — for anything bigger you’re pushed to Premium at $7/month. We have a fuller review of their privacy posture in Is iLovePDF safe? A 2026 privacy review.
Smallpdf
Swiss company, similarly strong compliance story (ISO 27001, GDPR, CCPA). The friction is the 2-document-per-day cap on the free tier; beyond that you need a $9/month Pro subscription. Their compression engine produces marginally smaller files on photo-heavy PDFs than iLovePDF in our testing, but the difference is within margin of error.
Adobe Acrobat online
Adobe’s free online tools sit on the same infrastructure as Acrobat Pro. The compression quality is excellent and the result is highly predictable. The friction is that you have to sign in with an Adobe ID to use it, and Adobe’s tracking footprint is heavier than the alternatives. Best fit if you already live in Adobe’s ecosystem.
PDF24
German company (smallware.de), unusual in offering both an online tool and a free desktop application. The desktop app processes files locally on Windows — no upload — which is a meaningful privacy advantage if you can install software. The online tool uploads like the others. Their tools page is among the cleanest of the major brands, with fewer aggressive upsells.
imisspdf
In-browser by default — your file is never uploaded. The PDF processing engine runs as WebAssembly inside your browser tab, reads the file from your disk into memory, compresses it locally, and offers the result as a download. There’s no server-side step, which means no upload time, no retention policy to read, and no risk of your file being in scope for a server-side incident. The file size is bounded only by your device’s RAM (typically 1-5 GB on a laptop, 500 MB-1 GB on a phone). No signup, no ads on tool pages, no watermark, no daily limit.
The architecture isn’t marketing — it’s structural. A server-based tool cannot easily pivot to in-browser; their entire business depends on the server model (upload caps gate Premium, server queues are how they monetise). An in-browser tool can’t add server processing without abandoning its core promise. The two approaches serve different threat models.
Step-by-step: compressing a PDF with imisspdf
This is the workflow if you want to try it. Total time: about 20 seconds for a 20 MB file on a modern laptop.
1. Open the tool. Go to imisspdf compress PDF. The page loads in your browser like any normal page. Behind the scenes, the WebAssembly PDF engine is downloaded and cached the first time — about 4 MB. Subsequent visits use the cache and load instantly, even offline.
2. Drop your file. Drag-and-drop onto the page or use the file picker. The file goes from your disk into your browser’s memory. Nothing is uploaded. If you watch your network panel (devtools → Network), you’ll see no outbound POST request carrying the file.
3. Choose compression level. Three presets:
- Light: lossless only — re-encode streams, subset fonts, deduplicate objects. No image resampling. Use this when you can’t risk any visual change.
- Balanced (default): resample images to 150 DPI and re-encode at quality 80. Good for most use cases — files for email, online portals, internal review.
- Strong: resample images to 96 DPI and re-encode at quality 70. Good for screen-only distribution where small visual changes are acceptable.
4. Process. Click compress. The engine works through your file in your browser. On a 20 MB file, that’s typically 10-30 seconds on a modern laptop, longer on a phone. Your CPU is the bottleneck, not your network.
5. Download. The compressed file appears as a download link. The file is already in your browser’s memory — clicking download just saves it to disk. The original file remains untouched on your disk.
6. Leave the page. That clears the file from browser memory. There’s nothing on a server to delete because there was nothing on a server.
Practical tips for better compression results
A few habits get you better results regardless of which tool you use:
- OCR scanned documents first, then compress. An OCR pass adds a text layer underneath the scanned image. Once that’s in place, you can downsample the image layer aggressively because the searchable text is preserved separately. We cover this in How to OCR a scanned PDF.
- Match DPI to actual use. 150 DPI for office printing, 96-120 DPI for screen, 300+ DPI only for commercial print. Most files arrive at 300-600 DPI and don’t need it.
- Remove embedded thumbnails. Many PDFs ship with built-in page thumbnails for fast scrolling in viewers. Modern viewers generate these on the fly; the embedded ones are dead weight.
- Flatten form fields and annotations before compressing if you don’t need them interactive. Flattening replaces the form-field machinery with the rendered output and the size win on form-heavy documents is significant.
- Don’t compress twice. Re-compressing an already-compressed file rarely shrinks it further and often degrades quality. If you need a smaller result, go back to the source and re-export with stricter settings.
Deciding which tool to use
The honest framing: the compression itself is roughly equivalent across the major tools. The difference is in everything around it.
Pick iLovePDF or Smallpdf when you want the broadest feature set and don’t mind uploading non-sensitive documents to a third-party server. They’re both legitimate, compliant companies with good security postures and clean track records.
Pick Adobe Acrobat online when you already use Adobe Creative Cloud or Adobe Sign and the integration matters more than the privacy or pricing model.
Pick PDF24 desktop when you’re on Windows, can install software, and want server-free processing with the broadest feature set from a single brand.
Pick imisspdf when the document is sensitive, when you don’t want to sign up for anything, when you’re on a flaky network, when the file is larger than a free-tier cap, or when you just want the operation to happen without any third party in the loop. The compression result is comparable; the architecture is different.
The frame that works best across all of these: decide per document, not per tool. A marketing flyer can ride iLovePDF without anyone losing sleep. A signed contract should not.
Try in-browser compression
If you’re handling a document where the upload itself is the wrong trade, try imisspdf compress PDF →. No signup, no upload, no watermark, no file-size cap beyond your device’s RAM. The file stays on your device the whole time.
Frequently asked questions
The FAQ block at the top of this article covers the most common compression questions. For deeper coverage of related topics, see Compress PDF without losing quality: lossy vs lossless and How to reduce PDF size for email.
Sources
- Ghostscript: Optimizing PDFs
- Smallpdf: How to compress PDF to a chosen size
- Smallpdf: Compress scanned PDF without losing quality
- iLovePDF Security & Data Protection
- iLovePDF Privacy Policy
- Adobe Acrobat online compress PDF
- PDF24 compress PDF
- Similarweb — iLovePDF traffic Q1 2026
Use Compress PDF: Reduce file size while optimizing for maximal quality. No signup, nothing uploaded.
Frequently asked questions
Yes — but only with lossless techniques (stream re-encoding, font subsetting, removing duplicate objects, dropping orphan metadata). These typically shrink office PDFs by 20-40% with zero visible change. The bigger wins (50-90%) require lossy image resampling, where small visual changes are deliberately accepted in exchange for a much smaller file. Whether that counts as 'without losing quality' depends on your use case — a 150 DPI screen-display version of a 600 DPI scan looks identical on a screen but loses detail under a print loupe.
Three common reasons. First, the file was already optimized — most PDFs exported by modern Word or Google Docs are within ~5% of their minimum size and there's nothing meaningful to remove. Second, it's a text-only document — text streams are already deflate-compressed and re-running compression on them yields almost nothing. Third, the embedded images are already JPEGs at a low quality setting, so the compressor either preserves them (no change) or re-encodes them (visible degradation for almost no size win).
Lossless removes redundancy from the file structure without changing how the document renders — duplicate objects, unused resources, uncompressed streams, redundant fonts. The result is bit-for-bit identical visually. Lossy goes further by resampling embedded images to a lower DPI, re-encoding them with stronger JPEG quantization, or converting color images to grayscale. Lossy can shrink an image-heavy PDF by 70-90% but you're trading some pixel-level detail for the size win. The right choice depends on whether the file will be printed at high resolution or just viewed on a screen.
Depends on the tool's architecture. Server-based tools (iLovePDF, Smallpdf, Adobe, PDF24's cloud option) upload your file to their infrastructure, process it, and delete it after a retention window — typically 1-2 hours. For non-sensitive documents this is fine. For confidential material (contracts, payslips, medical records, IDs), an in-browser tool that processes the file locally without uploading is structurally safer because the file never leaves your device. imisspdf and PDF24's desktop app use this model.
150 DPI for office printing on standard laser printers, 96-120 DPI for screen and email distribution, and keep 300 DPI only if the document is going to a commercial print shop. Most scanned documents arrive at 300-600 DPI, which is overkill for anything other than archival or professional print. Downsampling from 600 to 150 DPI on an image-heavy file typically yields a 75-85% size reduction with no visible change on screen.
Related articles
Convert PDF to Excel (Tables to Spreadsheet)
Pull tables out of a PDF into editable Excel rows and columns. When it works well, when to expect cleanup, and how to do it free.
How to Convert Word to PDF (Keep Formatting)
Turn a .doc or .docx into a PDF that looks the same on every device. Why PDF beats sending Word, and how to convert for free.
How to Convert a PDF to JPG Images
Export PDF pages as JPG or PNG images — one per page or just the ones you need. Free, in your browser, nothing uploaded.