How it works
How the online editor processes a PDF: text extraction and OCR, Cloud, Local or rule-based detection, review, and an export that removes covered text.
The editor runs as a web app in your browser. Most of the work, including OCR, face detection, rendering and building the redacted file, happens on your device. The only step that can involve a server is AI detection in Cloud PII mode, which is the default.
1. Open
You open a PDF from your computer, Google Drive or Dropbox. The file is kept in your browser's storage (IndexedDB) so the editor can reopen it after a reload. Password-protected PDFs are unlocked in the browser after you type the password. See Open and download files.
2. Read the text
For each page the editor reads the PDF's own text layer: every word and its position. Pages without a text layer, such as scans and photos, go through OCR in the browser. By default the editor also reads text inside pictures on typed pages, for example a scanned ID pasted into a contract. See OCR settings.
3. Detect personal data
Three detectors can run on the extracted text:
| Detector | Where it runs | What it finds |
|---|---|---|
| Cloud PII (default) | Detection API; page text is sent | Names, addresses, organisations, IDs, medical conditions and more |
| Local NER | In your browser, model downloaded once | The same kinds of entities, without sending text anywhere |
| Rule-based | In your browser | Exact patterns: e-mails, phone numbers, IBANs, card numbers, SSNs, PESEL, dates, postal codes, URLs, countries |
Only one AI model runs at a time: Cloud PII or Local NER. Rule-based detection can run next to either one. Face and signature detection run in the browser too. Details are in Detection settings.
4. Review
Each finding becomes a redaction box with a type such as person, email or credit card. You switch types on and off, remove single findings, select extra text and draw your own boxes. Nothing is final until you save. See Review and edit redactions.
5. Export
When you click Save PDF, the editor builds a new file in the browser:
- Each page is rendered as an image with the redaction boxes burned in, so nothing under a box survives as text or vector graphics.
- A new invisible text layer is added so the result is still searchable and copyable. Any word more than half covered by a box is left out of that layer.
- Pages are written at A4 portrait size.
- The original document's metadata, annotations, form fields, attachments and hidden layers are not copied. The output is a new PDF, not an edited copy of the original.
The output is flattened
Because pages are re-rendered as images, form fields stop being editable and the file can be larger than the original. Keep your original file if you need to edit it again later.
6. Download
The file is saved as redacted-<original name>.pdf. Downloading needs a free account, except for the built-in sample document. Signed-in users' scanned pages are counted towards their monthly page allowance, see Usage and limits.
Next steps
Quick start
Get started with the PDF Redaction Online editor: open a PDF, let AI find personal data, review the highlights, preview the result and save the redacted file.
Open and download files
Ways to open a PDF in the online editor (drag and drop, Google Drive, Dropbox, sample file), unlock password-protected PDFs and save the redacted result.