Most People Redact a PDF or Image Wrong. Here Is How to Do It Without Leaving a Trace to Recover

Every year, a redacted court filing, a leaked government document, or a shared contract makes headlines for the wrong reason. Someone selected the “redacted” text and it was still there. Someone deleted the black box in a PDF editor and the name underneath was untouched.

The redaction looked complete. It was not.

This article explains how real redaction actually works, how to redact a PDF, an image, an Aadhaar number, or a face so the content is genuinely gone, and what GDPR and India’s DPDP Act require before you share a document.

Two Ways “Redacted” Documents Still Leak Data

Every year, someone shares a document they believed was safe and it turns out the “redacted” text was still there underneath. It has happened to government agencies, law firms, and news organizations that used a highlighter tool or a drawn rectangle instead of real redaction.

Real redaction has one requirement that a cosmetic overlay does not meet: the original content has to be gone from the file, not covered by something on top of it. If the sensitive text or image data is still present anywhere in the file’s structure, it is not redacted, it is hidden, and hidden is not the same as removed.

Why How You Redact Matters as Much as What You Redact

There is a step people skip when they think about redaction: what happens to the file during the redaction process itself. Most free online redaction tools work by uploading your document to a server, processing it there, and sending back a result. That means the sensitive document, the one you are trying to protect, travels over the internet and sits on a server you do not control before it has been redacted at all. If your goal is to protect a document containing a medical record, a bank statement, or an identity number, uploading it to a third party before you have even removed the sensitive part defeats the purpose.

This is the specific reason ZeroCloudPDF’s Redact tool runs entirely inside your browser. The PDF or image never leaves your device, not during upload, not during processing, and not during download. There is no server in the process to intercept, store, or accidentally retain your file. You can turn off your internet connection after the page loads and the tool still works, because there is nothing left for it to send anywhere.

How ZeroCloudPDF Actually Redacts a PDF or Image

Here is the technical explanation, because how a redaction tool works determines whether it is trustworthy.

When you draw a rectangle or a circle over a page, the tool does not attempt to delete the underlying text from the PDF’s content stream, the internal data structure that stores selectable text and vector graphics. Deleting individual characters from a content stream while leaving the rest of the page intact is fragile and has repeatedly been the root cause of real redaction failures in legal and government documents, because fonts, character positioning data, and embedded metadata can retain fragments of the original text even after an edit.

Instead, every page of the PDF is rendered into a flat image, exactly as it would appear printed on paper, with your drawn shapes baked permanently into the pixels at full opacity. That flattened image then becomes the new page. The original text layer, the searchable and selectable data that made the PDF a PDF rather than a scanned picture, does not exist anymore on that page. There is nothing to search, nothing to select, and nothing to extract, because the words themselves have been replaced by a picture of the words with the sensitive part painted over.

The tradeoff is honest and worth stating plainly: the rest of the document, not only the redacted area, loses text selection and search after this process, and the resulting file is larger than the original. That is the cost of redaction that cannot be reversed. Any tool that promises to redact only the marked area while leaving the rest of the document fully searchable is, by definition, keeping the original text layer intact somewhere in the file, which is the exact vulnerability real redaction exists to close.

For images, the same principle applies without the extra step. The pixels you cover are permanently overwritten in the exported file. There is no original layer underneath to recover, because a redacted image from this tool is a new flattened image, not the original image with something drawn on top of it.

Step by Step: Redacting a PDF

1. Open the Redact PDF tool and select your file. Nothing uploads, the file loads directly into your browser’s memory.
2. Choose rectangle or circle from the shape toolbar.
3. Draw over any text, number, signature, or image you want permanently removed. Repeat across as many pages as needed.
4. Click a shape to resize it by dragging a corner, move it by dragging its body, or delete it if you placed it wrong.
5. Once every sensitive area is covered, click Generate Redacted PDF. The tool flattens every page and the finished file downloads straight to your device.

Always zoom in and review each page before sharing the output. Redaction is permanent, so it is worth thirty extra seconds to confirm nothing was missed, particularly text that continues onto a second line or a number that wraps across a page edge.

Step by Step: Redacting a Photo, Screenshot, or Scanned ID

The same tool accepts JPG, PNG, HEIC, WEBP, GIF, BMP, TIFF, and SVG files directly, so a photo of a document or a screenshot does not need to be converted to PDF first. Upload the image, draw over whatever needs to disappear, whether that is a face, a signature, an account number, or a barcode, and export. The process and the guarantee are identical to redacting a PDF: the covered pixels are gone from the exported file, not hidden beneath a layer.

Redacting an Aadhaar Number the Right Way

Aadhaar numbers come up constantly in KYC documents, rental agreements, job applications, and loan paperwork, and India has specific, binding rules about this that go beyond general privacy law. UIDAI’s masking rule requires that the first eight digits of the twelve digit Aadhaar number be covered before a copy is stored or shared, with only the last four digits left visible. This requirement has been reinforced through sector circulars, including RBI’s guidance for regulated entities accepting Aadhaar for KYC and IRDAI’s 2019 circular for insurers, both of which point back to masking the number before the document is used or filed.

To redact an Aadhaar card correctly, draw a rectangle over the first eight digits of the number specifically, leaving the last four visible if you need the document to still serve as a partial identity reference, or cover the entire number if no part of it needs to remain visible. The same applies to the QR code printed on newer Aadhaar cards, which encodes the same data in machine readable form and should be covered whenever the number itself is being masked, since scanning the QR code defeats the purpose of masking the printed digits.

Redacting Faces and Other Biometric Identifiers

A face in a photograph is personal data under most privacy frameworks in the same way a name or an ID number is, because it can identify a specific person. GDPR treats biometric data used for identification purposes as a special category under Article 9, which carries stricter handling requirements than ordinary personal data. If you are sharing a document that includes a photograph, a scanned ID with a face on it, or a screenshot containing people who have not consented to being identified, covering the face is not optional caution, it is the same category of requirement as covering a name or a number.

Draw a circle or rectangle directly over the face, adjust it to cover the full outline including hair and jawline if precision matters for your use case, and export as you would for any other redaction. Because the export flattens the image rather than overlaying a shape, the original facial data does not persist anywhere in the output file for later recovery through image editing.

What GDPR and DPDP Actually Require

Neither law uses the word “redaction” directly, but both are built around the same underlying idea: personal data that has been properly and irreversibly removed falls outside most of the law’s obligations, while data that has only been disguised or partially hidden does not.

Under GDPR, Recital 26 sets the standard for anonymization: for data to be considered anonymous rather than personal, the individual must no longer be identifiable, even indirectly, using any means reasonably likely to be used. A redaction that can be reversed, whether by deleting an overlay box or recovering an underlying text layer, does not meet that bar and the data is still treated as personal information subject to the full regulation. Article 9 separately singles out biometric data, health data, and other special categories for stricter treatment, which is why faces and medical details deserve particular care before sharing. Article 17 gives individuals the right to have their personal data erased on request, a right that proper redaction is often used to fulfill when full deletion of a document is not practical.

India’s DPDP Act, 2023, notified along with its Rules in November 2025, applies to digital personal data, defined around information relating to an identifiable individual. The Act does not expressly exempt anonymized data, but relying on its stated scope, personal data that has been irreversibly anonymized so that no individual can be identified from it is understood to fall outside the Act’s core obligations, mirroring the GDPR approach. As with GDPR, the operative word is irreversible. A document with a movable black box over a name is not anonymized under either framework, because the underlying data is still present and recoverable.

A Pre-Share Checklist

Before sending any document that contains personal or sensitive information, whether by email, upload, or messaging app, it is worth running through a short list.

• Confirm the redaction tool flattens or truly deletes content rather than layering a box on top of it.
• Open the finished file and try to select or search the area you redacted. If any part of it highlights or returns in a search, it was not removed.
• Check for the same sensitive data appearing elsewhere in the document, such as a name repeated in a footer or an ID number printed twice.
• Cover QR codes and barcodes if the data they encode matches something you redacted in text form.
• Redact before you upload the file anywhere, not after, if the tool you are using is not fully browser based.

Redaction is not something to get approximately right. A document that is ninety percent redacted with one missed line is functionally not redacted at all, because the sensitive information is still recoverable by anyone who reads it carefully. Taking the extra minute to use a tool that removes content rather than covering it, and reviewing the result before sending it, is the difference between a document that is actually safe to share and one that only looks that way.