Remove reviewed content through a new page-image document
Redact PDF takes a different approach from drawing a rectangle on top of an existing PDF. It renders every page, bakes your reviewed solid regions into those pixels and creates a fresh image-based PDF. The original text streams and interactive structures are not copied into the new document. Use it when a PDF delivery needs reviewed visible areas removed and you accept the consequences of flattening.
Choose a local PDF, inspect its pages and mark the regions deliberately. Automatic findings can help direct attention where supported, but they are suggestions rather than a complete discovery of every sensitive detail. The original stays unchanged, so keep it privately when provenance, editable text or another legitimate record requirement matters.
Review every page before confirming the plan
Searchable PDFs can expose extractable text to supported privacy-pattern checks. Scanned pages can use optional local printed-English OCR, and page-specific QR checks are available. Handwriting, unusual typography, other languages, faint text and complex layouts can fall outside reliable detection. A page with no finding has not been declared harmless.
Look for repeated names, identifiers, headers, footers, signatures, photographs and background content. A value redacted in a paragraph may still appear in a table or a scanned attachment. Review the complete document rather than only following suggested boxes.
Use solid regions with enough padding around the sensitive content. Letter edges, shadows and adjoining punctuation can extend beyond an obvious center. The goal is to cover the complete visible region, not to draw the smallest possible rectangle. A missed digit or repeated occurrence can matter more than whether the box looks elegant.
Every page is rebuilt, including unmarked pages
The export model renders the whole document rather than leaving original page objects under a cover. Marked solid regions become part of the page images, and fresh images are embedded into a new PDF. Unmarked visible content still remains in those page pictures, so the rebuilding process does not excuse an incomplete review.
Pages retain their MediaBox, CropBox, rotation and physical dimensions within the supported workflow. Standard requests 150 DPI equivalent and High requests 225, reduced when necessary for large pages. These are bounded rasterization choices, not a guarantee that every original vector detail can remain equally sharp at arbitrary magnification.
Photographic or densely detailed pages can increase the output size. The fresh PDF uses compressed PNG page images to preserve opaque redaction pixels. A larger output is therefore possible even when the source PDF was compact. Do not assume that a redaction tool is also a document compressor.
Know what flattening intentionally leaves behind
The new image-based PDF does not retain selectable or searchable text, live forms, links, annotations, comments, bookmarks, attachments, optional-content layers, actions or digital signature objects as original document structures. That is part of this output model, not a minor cosmetic change.
A recipient needing an accessible, searchable or interactive document may require another approved redaction workflow. Do not describe the result as preserving every capability of the source. PDF/A compliance, vector sharpness and other specialized properties can change, and no universal archival or accessibility certification is implied.
A visible signature mark may remain on a page image if you do not redact it, but it is not a cryptographic signature for the new file. Creating a redacted copy does not preserve the original digital signature's validity. Retain original signed files where necessary and follow the recipient's document requirements.
Verification is specific to the actual exported copy
The tool reopens the generated bytes, checks supported text and structural conditions and renders marked areas to verify solid coverage. Requested OCR or QR checks repeat where meaningful. Read unavailable or incomplete states rather than counting a created file as proof that every requested check succeeded.
This verification concerns the selected regions and supported checks. It does not establish that every unmarked part of every page contains no private information. A photograph in an appendix, a handwritten note or a recognizable location may need judgement that the detectors do not provide.
After downloading, open the actual saved PDF and inspect every page. Check the marked areas, their margins and the rest of the content. Make sure the file you send is the rebuilt copy rather than the original still open in another tab. Similar names can undermine a carefully reviewed redaction if the wrong attachment is chosen.
Budgets prevent an unbounded rendering job
Configured input admission is 20 MiB and 200 pages; encrypted PDFs are unsupported. Rendered page images have a 20 MiB budget, the raster workload is capped at 400 megapixels and the plan supports up to 500 marked regions. These constraints apply together with page and browser-memory checks.
A compact PDF can describe large or complex pages, so stored file bytes are not a complete estimate of work. Higher resolution needs more pixels and can hit a budget sooner. A refusal or reduced render scale should be understood from the reported limits, not bypassed by renaming the source.
Creative content and findings stay in the browser processing path. Application code, OCR and PDF resources still need to load from the site when required. This is local processing, not an assertion that a device is immune to extensions, other users or exposed downloads.
Complete a defensible handoff
Keep the original privately where appropriate, review the rebuilt file and confirm the page count, orientation, readability and intended covers. A successful download request is not proof that saving finished or that a client received the document. Describe the result as a reviewed image-based redacted copy, with the limitations that matter to its recipient.
For supported property inspection, PDF Metadata Viewer & Remover answers a separate question. For a rendered visual copy of selected output pages, PDF to Image can help with another local review. The redaction guide explains why an opaque flattened cover and a cosmetic blur should not be treated as equivalent protection.