How to Compress a PDF on Mac Without Uploading It Anywhere
TL;DR
macOS can already compress a PDF: Preview, File → Export, Quartz Filter → Reduce File Size. It is free, offline, and installed. It is also uncontrollable — one fixed setting, no target size, and on an already-compact PDF it often comes out *larger*. The job people actually have is not "smaller", it is "under 10 MB, because the portal says so", and that needs a tool that can aim. Dhito does it on-device: give it a byte budget, it finds the resolution that fits, writes a new file beside the original, never overwrites, and nothing is uploaded. $4.99/month after a 14-day trial, or $99 once.
There is a specific kind of small defeat in modern computer use. You have a 34 MB PDF. The portal accepts 10. You search "compress pdf", land on a site that wants the file uploaded, hesitate for exactly one second over the fact that it is a signed contract, and then upload it anyway, because you have a deadline and the alternative is unclear.
The alternative is not unclear. It is just badly documented. Your Mac has shipped with a PDF compressor for over twenty years, there are three distinct offline approaches, and only one of them can do the thing you actually need — which is not "make it smaller" but "get it under ten megabytes".
First, the free one you already have
Open the PDF in Preview, choose File → Export, and in the export sheet open the Quartz Filter menu and pick Reduce File Size. Save. That is the whole procedure, it is offline, it costs nothing, and for a photo-heavy document it often works well enough to end the problem right there.
Try it first. Genuinely. Three things it will not do, though:
It has no target. There is one setting, baked into the filter, and you get whatever it produces. If that lands at 12 MB against a 10 MB limit, your options are to run it again on the output — which compounds the damage — or to go looking for another tool.
It sometimes makes the file bigger. The filter downsamples embedded images and re-encodes them as JPEG. On a text-heavy PDF exported from Word, Pages or LaTeX there is little image data to remove, and the re-encoding adds overhead. People assume a compressor cannot make a file larger; it can, it does, and Preview will not warn you.
Its quality settings are aggressive. Reduce File Size was tuned for screen viewing. Run it over a scanned document and the result is legible but soft, and text that would have survived OCR sometimes does not.
There is a power-user extension of this: ColorSync Utility can duplicate the Reduce File Size filter and let you edit its image-sampling dictionary, giving you a gentler or harsher custom filter that then appears in Preview’s export menu. It works. Almost nobody does it, because you are hand-editing resolution and compression-quality values in a utility most Mac users have never opened, and even then you still have no target size — you have a second fixed setting instead of one.
The real job is a number, not an adjective
This is the part most compression articles skip. Nobody wants a "smaller" PDF. They want one under a specific number, because something on the other side of the transfer has a limit:
- Gmail stops at 25 MB per message.
- Court e-filing systems, visa and immigration portals, grant submission systems, insurance claim forms and job application sites commonly cap uploads at 5 or 10 MB, sometimes per document.
- Internal systems inherit whatever limit somebody set in 2014.
Against a hard number, a fixed-ratio compressor turns into a guessing loop: compress, check the size in the Finder, still too big, compress the output again, watch the scan turn to mush, give up, upload it to a website. The loop is the reason the upload sites win, not the compression quality.
Why everybody ends up on an upload site anyway
It is worth being fair about those sites, because the usual scare-story framing is wrong and people can tell. iLovePDF is ISO/IEC 27001:2022 certified, GDPR compliant, encrypts files in transit and at rest, and deletes uploads within two hours. Smallpdf makes comparable commitments. These are not cowboys, and for most documents on most days nothing bad happens.
The objection is structural rather than moral. Compressing a PDF is arithmetic on bytes you already have. It needs no server, no account, and no network. Routing it through one means:
- A document you control becomes a document in transit, briefly at rest on infrastructure you do not own. For a conference poster, fine. For a client bundle, a medical letter, a passport scan, a payroll file or anything covered by GDPR, HIPAA or the EU AI Act, you have created a processing event somebody may later ask you to account for.
- It does not work on a plane, on hotel wifi that blocks large uploads, or on the machine where the VPN policy forbids it.
- Free tiers meter you — two files an hour, a watermark, an account prompt at the worst moment.
- And you are uploading a file *to make it smaller*, which means the upload is the slowest part of the operation.
The same logic applies to every verb in this family — converting an image, stripping metadata, merging two documents — which is why it is worth solving once rather than per task. The broader version of this argument is in why cloud AI tools are a privacy risk.
Doing it locally, with a target
Dhito ships around 45 local file tools, and two of them are about PDF size. You reach either one by right-clicking a search result, selecting several files and using the selection bar, opening the Tools tab, or simply typing compress into the search box — the verb is a query like any other.
Compress gives you the choice Preview hides:
- Keep text selectable re-saves the document with its internal objects packed into compressed streams and strips the metadata fields — title, author, subject, keywords, producer, creator — which are small but free to drop and often carry boilerplate from whatever produced the file. Nothing visible changes. The text stays text.
- Flatten to images renders every page at a resolution you choose, from 50 to 300 dpi, defaulting to 120, and re-encodes as JPEG. This is what actually moves the number on a scan-heavy or slide-heavy document.
Either way, the result row reports the real before-and-after. If the file came out larger — which, as above, genuinely happens on already-optimised documents — it says the file was already compact rather than dressing up a failure as a success.
Compress to size is the one that ends the guessing loop. Give it a budget, five megabytes by default, and it works out the resolution and quality that will bring the document under it, flattens at those settings, and reports the size it achieved and the dpi it used. If the budget is unreachable even at the floor, it tells you that instead of quietly handing back something that still fails.
Four guarantees behind both, which matter more than the compression ratio:
1. Nothing is ever overwritten. report.pdf becomes report-compressed.pdf beside the original. If that name is taken, the next free one is used. If the folder cannot be written to — a read-only volume, somebody else’s directory — the result goes to Downloads and the row says it was redirected. 2. Undo is the Trash. Not a private history inside the app. The Finder’s own Put Back works, whether or not Dhito is running. 3. Bulk runs report honestly. Nine files are one job with one progress row, processed one at a time. Two failures name the two files and keep the other seven. 4. Nothing is uploaded. There is no provider to configure, no account, and no network path to audit. The tools work offline, on a plane, on a locked-down machine — and on a file you never indexed, because you can drag a PDF onto the window or paste it in and act on it immediately.
Which mode, when
| What you have | Use | Why |
|---|---|---|
| A text PDF from Word, Pages or LaTeX that is somehow enormous | Keep text selectable | The weight is structure and metadata, not pixels — and the text stays searchable |
| A scan, or a deck that is mostly screenshots | Flatten to images, 120 dpi | The weight is the images; downsampling is the only thing that moves the number |
| A hard limit: "under 5 MB" | Compress to size | It is the only one that aims |
| A contract you will need to search, quote or OCR later | Keep text selectable — never flatten | Flattening turns words into pixels |
| A document that has to stay accessible | Keep text selectable | Rasterising destroys the text layer and the tags that screen readers rely on |
| A photo album for a relative | Preview’s Quartz filter | It is free, installed, and good enough |
The caveat worth repeating
Flattening is a one-way door. Once a page is a JPEG, the words in it are pixels: no selection, no copy, no Cmd-F, no screen reader, no text layer for anything downstream to read. If the document is going out for signature, into a case file, or anywhere it has to remain searchable, take the smaller saving from the lossless mode and live with the size.
You will still have the original to fall back on, because nothing was overwritten. That is not a small detail — it is the difference between trying the aggressive setting and not daring to.
And macOS 26 Tahoe’s Quick Actions?
Tahoe gave Spotlight its largest upgrade ever, quick actions included — rotate a thing, convert an image, create a PDF. They are useful and the instinct is exactly right. Compression is not among them, and neither is a size budget. The full breakdown is in the Dhito and Spotlight comparison.
Compressing a PDF is a small task. It is also one most people do dozens of times a year, usually in a hurry, and usually with a document they would rather not hand to a stranger to save four megabytes. Preview handles the easy half for free. For the half that comes with a number attached, the tool should aim — and it should never have needed the internet to do arithmetic on a file already sitting on your disk.
Frequently asked questions
Can macOS compress a PDF without extra software?
Yes. Open the PDF in Preview, choose File → Export, and pick Reduce File Size from the Quartz Filter menu. It costs nothing, runs offline, and is already on the machine. What it will not do is let you choose how hard to compress or aim at a size limit — the filter has one fixed set of sampling settings, and you either accept the result or you do not.
Why did my PDF get bigger when I compressed it?
Because compression is not one operation. Preview’s Reduce File Size filter downsamples the images inside a PDF and re-encodes them as JPEG. If the document was mostly text, or the images were already well compressed, re-encoding adds bytes rather than removing them and the export comes out heavier than the original. This is normal and it catches people out constantly. A tool that reports the real before-and-after rather than assuming success saves you from sending the wrong file.
Is it safe to use iLovePDF or Smallpdf for confidential documents?
They are more careful than people assume — iLovePDF is ISO/IEC 27001:2022 certified, GDPR compliant, encrypts in transit and at rest, and deletes uploads within two hours. The objection is not carelessness, it is shape: your document leaves your machine and is processed on someone else’s infrastructure. For a holiday photo album that is irrelevant. For a signed contract, a passport scan, a medical report, a client’s discovery bundle or a payroll run, you have turned a local chore into a data transfer — one that, in a regulated environment, you may have to be able to describe to somebody. The upload is a risk surface that does not need to exist, because the compression itself never needed a server.
How do I get a PDF under a specific size, like 5 MB?
That needs a tool that takes the number as input rather than one that compresses by a fixed amount and leaves you to check. Dhito’s Compress to size tool takes a byte budget, works out the resolution and JPEG quality that will fit under it, flattens the pages at those settings, and then reports the size it actually achieved and at what resolution. If even the lowest settings overshoot, it says so plainly instead of handing you a file that still fails the upload.
Does compressing a PDF lose quality?
It depends which mode you pick, and the difference matters more than the file size does. Keeping text selectable re-saves the document structure and drops metadata — nothing visible changes and the text stays searchable. Flattening to images rasterises every page at a chosen resolution, which is what genuinely shrinks a scan-heavy file, but it turns words into pixels: no selection, no copy, no Cmd-F, no screen reader, no OCR layer. Never flatten a contract you will need to search or quote from later.
Can I compress several PDFs at once, and what happens to the originals?
Select as many as you like and run the tool once — they are processed one after another as a single job with one progress row, and if two of nine fail, the run names those two and keeps the other seven. Nothing is ever overwritten: each result is written next to its original with a -compressed suffix, so report.pdf becomes report-compressed.pdf. If the source folder cannot be written to, the result goes to Downloads and the row tells you it was redirected. Undo puts the new files in the Trash, which means the Finder’s own Put Back works whether or not Dhito is still open.
Related Articles
How to Chat with a PDF Locally on Mac (100% Offline & Private)
Want to ask questions to a 100-page PDF without sending your data to the cloud? Discover how local AI tools like Dhito make private document Q&A easy and fast.
Why Cloud AI Tools Are a Privacy Risk (And What to Use Instead)
Thinking twice before uploading that contract to a cloud AI? Learn about the privacy risks of cloud-based LLMs and the shift toward on-device processing.
Chat with Your Files: How Local Document Q&A Works Offline
Ask questions across your own documents, contracts, and research reports and get answers with citations — with every page staying on your Mac. Here is how local document chat works.
Want to try Dhito?
Download Dhito and experience the power of local semantic search today.