Black out text in a PDF with a box drawn over it, and the words are often only hidden on screen: they stay in the file, and anyone can select them, copy them or find them by search. PDF redaction should take them out. So we made two small PDFs, blacked out the same lines with four free tools that work in a browser, and checked what each saved file still held.
The test
- The files: a two-page contract and a one-page bank statement, every line marked SPECIMEN, written with a PDF library rather than with any of the tools tested. Five lines hold a secret (names, an address, account numbers), and six lines are to be left alone.
- The tools: LensUp’s PDF editor, and the Redact PDF tools of iLovePDF, Smallpdf and i2PDF, each on its own site, in Chromium on a desktop, with no account, on 13 October 2026. In each, we marked the whole of every secret line with the tool’s own redaction tool (LensUp’s is called Cover), dragging across it, and saved the file the way the tool offers.
- LensUp, both ways: when you save a file with covers, LensUp’s editor asks whether to remove what is under them. We answered «Remove it», and, as the control, «Keep as is»: a cover that only hides, which the test has to catch.
- The checks, by a script, on each saved file:
- Gone from the file: each secret line’s label, its whole value, and the value’s own words (those that mix letters and digits or have six characters or more, and are not in the lines left alone), looked for in every object and stream of the file a PDF library can read, decoded, and in the text a PDF reader lets you select.
- Hidden on the page: each page drawn at 72 dpi; in a secret line’s box, the share of dark pixels must fall outside what a line of text gives here, 2 to 40 per cent: paper is below that, a black bar above it.
- The other lines: still drawn on the page, their dark share within that range, and still text you can select.
Results, as of 13 October 2026
| Tool, and what we pressed | Secret lines gone from the file | Secret lines hidden on the page | Other lines still on the page | Other lines still text you can select | Size of the two files |
|---|---|---|---|---|---|
| LensUp: Cover, then Remove it | 5 of 5 | 5 of 5 | 6 of 6 | 0 of 6 | 225.1 KB |
| LensUp: Cover, then Keep as is (the control) | 0 of 5 | 5 of 5 | 6 of 6 | 6 of 6 | 6.9 KB |
| iLovePDF: Redact, then Download file | 5 of 5 | 5 of 5 | 6 of 6 | 6 of 6 | 6.4 KB |
| Smallpdf: Redact, then Download | 5 of 5 | 5 of 5 | 6 of 6 | 6 of 6 | 4.7 KB |
| i2PDF: Redact, then Start | 5 of 5 | 5 of 5 | 6 of 6 | 6 of 6 | 60.5 KB |
What it means
Every tool’s redaction took the secret lines out of the file and hid them on the page. The control shows why both checks matter: LensUp’s cover alone hid all five lines on the page and took none of them out of the file.
Where the tools differ is the rest of the document. iLovePDF, Smallpdf and i2PDF kept the other lines as text you can select. LensUp’s «Remove it» saves every page as a picture: the other lines still show, but none of them can be selected, copied or found by search, and the two files grew from 4.3 KB to 225.1 KB. That is where LensUp did worse here.
Why LensUp works that way
We first built a remover that cuts only the covered words out of the file, and our own reviews kept finding kinds of PDF in which covered words could survive it. So «Remove it» keeps nothing of the file but pictures of its pages, drawn at up to 200 dpi (a very large page coarser), for files of up to 100 pages; links and form fields go too. Not a word of the original text, not even the file’s title, was left in what it saved here.
Which one to use
If the rest of the document must stay text, to search it or copy from it, iLovePDF, Smallpdf and i2PDF each did that with these files. If you would rather nothing of the original text were left in the file, LensUp’s «Remove it» does that, at the cost of text you can select. Whether a tool uploads your file is a separate question this test did not check; for LensUp’s PDF editor, we measured that separately.
Check it yourself
- All results (CSV): one row per tool, way and file — each check’s count, the pages, the file’s size, and where and when it ran.
- Every saved file (ZIP): what each tool saved; LensUp’s run record, with its site and build; and for the other tools, the page and the steps of each run.
- The files we made: the contract, the bank statement, and where each line is (JSON).
Limits
- Two small files, plain text in a standard font. Real documents can hold words where these do not, such as form fields, comments, bookmarks, attachments and text inside pictures, and a tool that passed here may miss them there.
- One run per tool and file. Before the runs we opened each tool with the same files to learn its steps. i2PDF’s contract ran twice: in our first run its Redact tool went back to Select after the first line, which our run did not expect, so it marked one line of three. That run is not in the results; its file is in the ZIP.
- Not tested: finding words to redact by search, the tools’ own finding of names and numbers, scanned pages, and whether a tool uploads the file.
- Not run: pdfFiller, which needs an account; Xodo, whose free use is for personal needs; and PDF24, which we left out: its box tool puts each box in the middle of the page, to be moved and resized.
- We built LensUp and chose the test. The files are above for anyone who wants to rerun it.