Local or cloud?
“Runs in your browser” is a concrete architecture claim, not a decorative privacy phrase. Verify the data path before choosing a compressor for a sensitive document.
Ask where the bytes go
A browser page can process a file locally, upload it to a server, or do both. A “secure” label does not answer which path is used. Look for a precise statement about upload behavior, storage, deletion, subprocessors, and analytics.
Local processing reduces server exposure, but the website host still receives ordinary page requests and network providers still process connection metadata. It is not the same as operating fully offline.
Separate content from telemetry
A privacy-conscious tool should not place filenames, file sizes, target values, page text, or document contents into analytics events. Page-count analytics and task outcomes can be measured without copying the document itself.
Session replay must be consent-based and configured so file inputs and typed content are not captured. A separate privacy-preserving page counter can provide an all-visitor denominator.
- ■File bytes stay in the tab
- ■No filename in analytics
- ■No document text in events
- ■Session replay requires consent
Match the document risk
Identity, immigration, tax, employment, school, legal, and medical documents deserve a stricter review than a public brochure. If the tool’s boundary is vague, do not upload the file.
UnderLimit PDF performs the compression in the browser. It does not retain files, but rasterization can remove document features. Privacy and fidelity are separate checks; both must pass before submission.
Put the limit first
Run one local compression pass, then inspect every page before submitting.
Open the compressor →