Why PDFs get so large in the first place
A PDF is a container, and its size is almost always decided by the pictures inside it. When you scan a document at 600 DPI in colour, the scanner produces a full-page colour photograph — roughly 25 MB of raw pixels for a single A4 sheet. Ten of those pages make a 100 MB PDF that is impossible to email.
Phone photography makes it worse. A modern phone camera produces a 12-megapixel JPEG for every page, complete with shadows, background clutter and a slight tilt. Five pages photographed on a phone can easily exceed 15 MB, even though the information content — black text on white paper — could be stored in a fraction of that.
There is a second, quieter cause. Many PDF writers duplicate embedded fonts and images, or store uncompressed streams. A text-only PDF created by a word processor can arrive at 8 MB because a single logo image was inserted eleven times at full resolution.
What each compression level actually does
Light rewrites the file without touching its page images. Unused objects are dropped, streams are recompressed and metadata is trimmed. The typical saving is 5–20%, and crucially your text stays selectable and searchable. Use it when the file is only slightly too big.
Recommended renders every page and re-encodes it as a JPEG at 150% resolution with 72% quality. This is what the big online compressors do. Expect 60–85% smaller files, at the cost of text becoming part of the page image rather than a selectable layer.
Strong drops to about 105% resolution and 50% quality, which is still perfectly readable for forms, receipts and statements. Extreme goes to 80% resolution and 40% quality — useful when a portal has a brutal size limit and legibility matters less than fitting.
Target size is different: instead of a fixed recipe, the tool samples a few pages, measures how much each quality setting would produce, extrapolates to the whole document, picks the best setting that fits your target, and then verifies the real result and refines it if needed.
How to choose a level for a real document
For a scanned application form that must stay sharp enough for a clerk to read: Recommended, with grayscale switched on if the form is black and white anyway. Notices the difference in size immediately.
For a presentation or design file where colours and gradients matter: use Light. Compression that rebuilds pages flattens vector art into pixels, which is the wrong trade for something that needs to look crisp.
For a bank statement or invoice with small print: Strong is usually the sweet spot. Text at 105% resolution stays legible in 50% JPEG quality because the original is high-contrast black on white.
For a portal that says “max 200KB” and refuses to tell you anything else: use the dedicated Compress PDF to 200KB page, which searches for the best quality under that number instead of making you guess.
What you lose, and when it matters
Be aware of one honest trade-off. The levels that produce dramatic savings rebuild every page as an image, which means the text layer disappears. After that, the PDF cannot be searched with Ctrl+F, text cannot be selected or copied, and screen readers cannot read it aloud.
That is usually a fair trade for a scanned form, where the original had no text layer anyway. It is a poor trade for a contract, a report or a textbook. When words matter, use Light compression, or keep an uncompressed copy alongside the compressed one.
Two smaller things also change: interactive form fields and annotations on pages get flattened into the image, and any internal links stop working. Again, that is normal for this kind of compression, and it is why every level is clearly labelled in the interface rather than hidden behind a single “Compress” button.