Key point. Nothing is uploaded. Files are opened with FileReader, PDF text is extracted locally, and the anonymized output is text you can copy or download.
The concrete problem
A document often contains more identifiers than the paragraph you intended to quote. Contracts, invoices and reports repeat names, dates, account details and internal references.
Formatting can hide meaning. A table cell, a header or a footnote may carry the one value that identifies the organisation, even if the main paragraph looks clean.
Different file types behave differently. Treating a scan like a text PDF, or expecting DOCX styling to survive, leads to false confidence.
Practical steps
- Choose the smallest file or excerpt that contains what you need.
- Open a .txt, .pdf with embedded text, or .docx file in the tool.
- Read the extraction note: long files, PDF limits and DOCX plain-text conversion are reported in the interface.
- Review highlighted findings and add free-text identifiers manually, especially from tables and footnotes.
- Copy or download the anonymized text. Do not expect a reformatted PDF or DOCX back.
How PromptScrubber helps
TXT-family files are read as plain UTF-8 text. PDFs expose only their embedded text layer, page by page. DOCX files are reduced to plain text; layout, images and table order are not guaranteed.
After extraction, the same detection and manual review workflow applies as for pasted text. Stable placeholders keep repeated references consistent.
Limits and responsibilities
- Scanned or image-only PDFs are refused because OCR is not implemented.
- DOCX formatting is not preserved; the export is a .txt file.
- Long files may take longer to process locally; oversized input is not silently truncated.
- The restore map contains original values and must be protected if exported.
- Text extraction can miss structure or reorder content; compare with the source before relying on it.
Frequently asked questions
Which file formats are supported?
Plain text formats such as TXT, MD and CSV, PDFs with an embedded text layer, and DOCX files. Other formats are refused with an explanation.
Can you OCR a scanned PDF?
No. OCR is not implemented. A PDF without usable text is rejected rather than processed incorrectly.
Will my DOCX layout survive?
No. The tool extracts plain text. Tables, styling and images are not preserved in the anonymized output.
Where does the file go?
Nowhere. It is read in the browser with FileReader and processed locally. There is no upload endpoint.
Prepare your text
Open the tool: paste, review and anonymize text without uploading it.