Recovery
Open the app: Repair a damaged PDF → (products.documentize.ai)
In one sentence
Recovery takes a PDF that will not open, opens only partly, or shows broken text, tries repairs from the gentlest to the most drastic until one passes your checks, then cleans hidden data from the result and hands you an "X-ray" report of everything it did.
What makes it unique
- A repair ladder, not a single trick. It starts with the least invasive fix and only escalates if needed: triage of the damage, rollback to an earlier revision of the file, a standard structural repair, rebuilding the cross-reference table, rebuilding the page structure, salvaging damaged content streams, recovering fonts, and finally reassembling content. It stops at the first rung whose result passes.
- You set the acceptance test. You can reject a repair that has fewer pages than you expect, or whose pages mostly have no readable text. A "repaired" file that is secretly missing half the document is not accepted.
- It tells you the truth afterwards. The report says which rung worked, which pages are original, repaired or reconstructed, what was dropped, and a confidence level.
- It also cleans. After a successful repair it removes hidden data under a named policy and re-scans its own output to prove it; anything it could not prove gone is named in the report.
- It salvages extras. Images, attachments, metadata and text/form values pulled out of the file come back as additional downloads — even useful when the PDF itself is beyond repair, as the site's page puts it.
- It tells you when the file was never a PDF. A renamed EPUB, Word file, image and so on is handed back under its real extension instead of a dead-end error.
When to use it / when not to
- Use it for a truncated download, a file that opens blank, "the document is damaged" errors, missing pages, text that cannot be searched because of broken fonts — or when you want to scrub hidden revisions and metadata from a file before sharing.
- Not for password-protected PDFs. Encrypted files are reported as unrecoverable, not guessed at. Unlock first.
- Not as a general "clean up" for healthy files you just want smaller → Compressor or Optimize.
- Not for hiding something visible (a name, a number on the page) → Redactor.
- Not a miracle. If almost nothing is left in the file, you get an honest error with the reasons.
Before you start
- One PDF at a time: the app works on a single file.
- Keep the original; if the repair is poor, a different run (different settings) starts from it.
- Decide whether you want cleaning. The default policy is "Full", which strips metadata, attachments, history and hidden content. That is great before sharing and surprising if you wanted your metadata or attachments kept. Pick None or Keep attachments accordingly.
Step by step
- Open the Recovery page and drop the damaged PDF.
- Open Advanced options if you need a stricter check or a different cleaning policy.
- Start the repair and wait ("Repairing PDF…").
- Download
repaired.pdfand any extra files (images, attachments, and so on). - Read the PDF X-ray report, especially the confidence level and which pages were reconstructed.
Options explained
| Option | What it does | Default | Recommended |
|---|---|---|---|
| Expected page count | Rejects a repair that ends up with fewer pages than this. 0 means no check | 0 | Set it if you know how many pages the file should have |
| Minimum text coverage (%) | Rejects a repair unless at least this share of pages has extractable text. 0 means no check | 0 | Set e.g. a high number for text documents; leave 0 for scans |
| Sanitization policy | What to strip from the repaired file, then verify by re-scanning | Full | See below |
| Include full diagnostic report (PDF X-ray) | Adds the detailed report to the result | on | Keep on |
Sanitization policies
| Policy | In short |
|---|---|
| Full (default) | History, metadata, attachments, active content, and content hidden in plain sight |
| Structural | Residue outside the page; page content untouched |
| Public release | Full, plus image EXIF metadata |
| Court filing | Strict: if anything is still detected, the run fails |
| Internal share | History and metadata only; annotations kept |
| Keep attachments | Structural, but embedded files stay |
| None | Skip sanitizing; return the repaired file as it is |
What you get
repaired.pdf, a new file; your upload is not changed.- Extra files salvaged from the original (images, attachments, metadata, recovered text or form values), when there are any.
- The report: whether it recovered, the rung that was accepted, page count versus expected, per-page provenance, dropped features, damage notes, font repairs, how the page order was decided, a confidence level with reasons, and — if sanitizing ran — a certificate of what was found, removed, deliberately left and why.
- If the upload was not a PDF, you get the same bytes back under the right extension and an explanation; nothing is repaired.
Quality & limits
- Not every file is saveable. If no rung produces an acceptable result you get an error that lists the damage notes and the last failure.
- Reconstructed pages are reconstructed. Check the report: pages marked repaired or reconstructed may differ from the original (missing content, rebuilt fonts).
- Sanitizing never costs you a successful repair. If the cleaner itself cannot run, you receive the repaired file unsanitized (and the report has no certificate).
- Stricter checks can cause failures. A high text-coverage setting on a scanned document will reject good repairs, since scans have no text.
- Proof of removal has limits. The site says what could not be proven gone is named in the report; read that section before you rely on it for legal use.
- This app handles one file per run and does not offer a continuation after a free-plan time stop.
- Results are kept for about two days. The site marks the feature as free during beta (verify on the live app).
Works well with
- Before: nothing — repair comes first. If you are not sure the file is damaged, try opening it again.
- After: Compressor, Optimize, Searchable (if fonts were lost), Protector, Sign.
- Recipes: Rescue a damaged PDF, Safe sharing.
Common mistakes & fixes
- "It said no recovery stage produced an acceptable result" → lower the expected-page or text-coverage gates, or read the damage notes in the message; the file may simply have too little left.
- "My metadata and attachments disappeared" → that is the default Full policy. Re-run with None or Keep attachments.
- "It says the file is encrypted/unrecoverable" → Unlock it, then run Recovery.
- "I got a .docx/.epub/.jpg back" → your file was not a PDF; open it with the right program or convert it.
- "Text is gibberish after repair" → check the report's font repairs; run Searchable to rebuild text from the page images.
Same thing elsewhere
Recovery runs in the browser and through the web API; no dedicated command-line plugin or MCP tool was verified for it. See Beyond the browser.
Try it
Take a healthy PDF and make a damaged copy: cut off the last few kilobytes with a file tool (or ask a colleague for a broken download). Upload it to Recovery with Expected page count set to the true number of pages and policy None. Read the X-ray: which rung worked, and were any pages marked reconstructed? Then run the same file with Full and compare what the certificate lists.