The black-box lie that keeps embarrassing professionals
Every few months another organisation learns this publicly: a legal filing, a report, a manifesto 'redacted' by drawing black rectangles in a PDF editor is released — and within hours, someone copies the text from beneath the boxes and publishes what was hidden. The rectangles were annotations floating over the content; the content itself, fully intact, travelled with the file. True redaction is not covering — it is removal. The sensitive content must cease to exist in the file, in every layer: text, metadata and structure.
Redactly performs the honest version. You mark regions on each page, and the tool re-renders every page into a fresh document with your boxes burned into the page image itself. In the output file there is no text layer under the black at all — the pages are pictures, and a picture of a black box contains a black box, not a secret. The verify mode then proves it in the only convincing way: it attempts text extraction on your redacted output, live, and shows you what an adversary would find.
Why re-rendering is the consumer-grade gold standard
Industrial redaction systems rewrite a PDF's content streams surgically, excising the characters and glyphs in the marked regions. Done perfectly, that preserves selectable text elsewhere. Done imperfectly — and the history of failures says imperfectly is common — fragments survive: a word split across operations, text stored twice, a thumbnail rendering cached in the file. Re-rendering trades away text selectability for a guarantee a non-specialist can verify: nothing that was on the page survives except what the rendered image shows.
For the redaction jobs ordinary people and small organisations actually face — a statement for a dispute, a contract for publication, records for a freedom-of-information release, screenshots and scans for a report — that trade is correct. The recipient needs to read the document, not mine its text layer; and you need certainty, demonstrable with the verify button, that the salary figure, the account number, the name, the address are not lurking underneath. Certainty beats elegance when the content is sensitive.
How to redact a PDF, step by step
Load the document and take page one slowly. Drag a box over each sensitive item — and be generous: cover the whole field including its edges, because descenders, punctuation and the last digit of a number love to peek from under a tight box. Work page by page with the page selector; the box counter tracks your coverage per page so nothing is half-done. Names hide in headers, footers, signatures, ID numbers in margins and file paths in print footers — check the furniture of every page, not just its body.
Press Build and the redacted copy is rendered at high resolution. Before you send it anywhere, press Verify: the tool attempts extraction on the output and shows the result. Then do the human check — open the redacted file and look at every page once, because redaction failures are ultimately visual: the box that missed the second line of an address, the total that appears again in a summary on page nine. Redact the document you have, not the one you remember.
What to hunt for: a field guide to hiding places
Direct identifiers are the obvious quarry — names, emails, phone numbers, account and ID numbers, addresses, dates of birth, signatures. Then hunt the second order: figures that identify by combination, such as an exact salary plus a job title in a small team; case numbers that unlock other systems; QR codes and barcodes, which are machine-readable identifiers people forget are content; photographs of faces and number plates; and the document's own metadata, which this tool's re-render discards entirely, since the output is built fresh from page images.
Context creates a third order: if the surrounding text says 'the complainant, a 34-year-old nurse from…' you may need to redact descriptors, not only names. Redaction is an editorial act with legal consequences, and the tool's role is to make the removal technically certain — deciding what must be removed remains your responsibility, ideally against whatever disclosure rules govern your situation. When in doubt, redact wider and check the governing guidance; over-redaction embarrasses no one.
Redaction versus its lookalikes
Three operations wear similar clothes. Annotation — drawing boxes in an editor — hides nothing and must never be trusted. Cropping trims a page's visible area but keeps the full content beneath the crop box, retrievable by anyone who resets it. Watermarking declares ownership while leaving content readable by design. Redactly's re-render is none of these: it destroys the underlying representation and substitutes a picture of the result. Knowing which operation a tool performs — by asking what an extraction attempt would find — is the entire literacy of safe disclosure.
That is why the verify mode is not a garnish. Run it on any tool's output, including this one's, before a sensitive release: extract the text, search for the strings you meant to kill — the name, the number, a distinctive word — and confirm their absence. A redaction you have not verified is a hope; a redaction you have verified, with the method shown, is a fact you can sign under.
Privacy: redaction material never leaves the room
Consider what an upload-based redactor receives: the unredacted document — the single most sensitive version of the file that will ever exist. Redactly's architecture avoids that paradox entirely. Your original is rendered and boxed inside your browser tab; the redacted output is generated on your device; nothing is transmitted, so there is no server copy of the pre-redaction file to breach, subpoena or mishandle.
Handle your own copies with the same discipline: the original remains sensitive forever — store it as you would the secret itself, distribute only the verified redacted copy, and never 'helpfully' send both. The tool guarantees the technical half of redaction. The procedural half — one file, the right file, to the right people — is yours, and it is the half that most real-world failures actually fail on.
Frequently Asked Questions
How is this different from drawing a black box in a PDF editor?
Editor boxes are annotations floating over intact text — anyone can copy the words from underneath. Redactly re-renders each page with your boxes burned in, so the covered content does not exist anywhere in the output file.
How can I be sure the text is really gone?
Use Verify mode: it attempts text extraction on your redacted output, exactly as an adversary would. Because pages are rendered images, extraction finds nothing beneath the boxes — the proof is shown to you live.
Is text still selectable in the redacted PDF?
No — pages become high-quality images, which is the mechanism that guarantees removal. Recipients can read and print normally; keep your original (securely) if you need the text layer.
What should I remember to redact besides names?
Account and ID numbers, contact details, signatures, QR codes and barcodes, faces and plates in images, identifying figure combinations, and content in headers and footers on every page — not only page one.
Does redaction remove document metadata?
Yes. The output is built fresh from rendered pages, so original metadata, hidden layers and embedded text do not carry over into the redacted file.
Is my unredacted document uploaded anywhere?
Never — that is the point of the design. The sensitive original is processed only in your browser and never leaves your device.