Before you use a sensitive document with AI

Prepare sensitive documents before using AI.

Upload a PDF, DOCX, or pasted text. DocRinse finds names, SSNs, addresses, and account numbers, then creates a sanitized draft you review before using AI.

Local text preflightPasted text gets a browser-side first pass before backend review.
No account requiredTry pasted text instantly. Pay only if you need full document processing.
Start low-riskUse a sample or low-risk document before testing private files.
Private alpha workflow

Built for review, not blind upload

Tax + banking first

Clean first. Review second. Then take the result wherever you choose.

Document type detectionIncluded
Private-detail detectionIncluded
Sanitized draft reviewIncluded
Structured extractionIncluded
Cleaned exportsPaid plans

How it works

Clean it, inspect it, then decide what happens next.

01

Upload or paste

Start with a PDF, DOCX, TXT file, or pasted document text for quick testing.

02

Review what was found

See likely names, SSNs, account numbers, addresses, and IDs before anything moves forward.

03

Export or use with AI

Download the result in the formats supported for that document type, or export the reviewed sanitized draft.

Why this exists

Most people want AI help with private documents, but they do not want names, IDs, account numbers, or addresses floating around in the raw prompt.

No raw-file LLM cleanup Browser-held cleaned copy Backend file review is disclosed

DocRinse cleans first. You review. Then decide what to send to AI.

Raw AI upload DocRinse workflow
Raw file sent straight to AISanitized draft created and reviewed first
No visibility into what was removedRedactions and counts shown explicitly before export
Hard to know what changedAudit record with hashes and settings kept for review
Trust based on promisesTrust based on visible control and inspectable steps

Current validation evidence

DocRinse is checked against public samples and synthetic privacy gauntlets before trust claims move forward. These are internal QA gates, not a guarantee that review is optional.

9/9public-domain prose excerpts preserved without private-detail false positives
7/7external public and mock PDFs passed with no watched private terms leaking after rinse
34/34synthetic adversarial privacy docs passed with zero answer-key term leaks

Who it's for

Built for narrow, real use — especially tax, banking, benefits, and client-facing document review.

IND

Individuals and families

Use AI to understand your tax return, bank statements, or account documents without uploading your SSN, address, or account numbers to an AI tool.

FIN

Tax and financial-document helpers

Clean client-ready text, Word docs, or text-based PDFs before taking them to an AI tool. Best fit today is tax and statement workflows, not every document under the sun.

OPS

Privacy-conscious operators

If your main need is “clean this before I paste it into AI,” DocRinse is a better fit than a vague promise and a blind upload button.

Clean a document

Remove private details before using this document with AI. Your original is processed once and is not stored as a review copy.

Monthly access active

DocRinse Plus workspace

20 documents/month • 200 pages/month • renews monthly

Billing cycleCurrent month
Next resetAt renewal
Access recovery

Restore this workspace with your checkout email and restore key. Keep the key somewhere boring and findable.

Documents used0 / 20
Pages used0 / 200

Start a new private rinse when you have another sensitive document. DocRinse keeps the workflow record privacy-safe instead of building a raw-document archive.

Cleaned version created
Private details detected
Redaction report ready
Cleaned text ready to use elsewhere

Usage updates after each completed Plus rinse. Refresh if you restored access on another browser.

1

Add your document

Paste text instead
Free Preview limit: 3,000 pasted characters. 0 / 3,000
2

Choose cleaning level

3

Choose access

Start free to review what DocRinse finds. Choose a paid rinse when you need exports.

Already paid or restoring access?
Normal checkout restores access automatically. Use these controls only if support directs you.
Upload a file or paste document text to start.
Checking support for scanned PDFs...

Common questions

Straight answers on what this does and what it does not do.

Why not just upload the document directly to ChatGPT or Claude?

Sensitive documents often contain private details you may not want in an AI prompt. DocRinse gives you a prep step first: find likely sensitive details, review the result, then decide whether to use the cleaned draft with AI.

Does DocRinse keep my documents?

Your original document is used to create one cleaned result. The backend returns the cleaned result to your browser and does not keep a plaintext review/export copy. Export any result or report you want to keep.

Why are some details still visible after cleaning?

DocRinse looks first for direct identifiers: names, SSNs, tax IDs, account numbers, routing numbers, full street addresses, dates of birth, and similar private details. The stricter cleaning setting removes more context by default because privacy matters more than preserving every detail. Use a lighter setting only when readability matters more, and always review the cleaned result.

How accurate is private-detail detection?

DocRinse helps find common direct identifiers in structured financial documents: SSNs, tax IDs, account and routing numbers, dates of birth, addresses, labeled IDs, and repeated name variants. Current internal QA includes public-domain prose, public sample PDFs, mock financial documents, OCR companion files, and synthetic adversarial privacy packs. You should still review the redaction report and cleaned output before export.

What document types are supported right now?

PDF, DOCX, and plain text uploads, plus pasted document text. The engine is tuned for tax returns (1040, W-2, 1099s), bank statements, brokerage statements, lending documents, and several benefits-style records. Support varies by file type. The available export options depend on the document and what can be generated and validated.

What about scanned PDFs or image-based files?

Scanned and image-based files can be harder to verify than text-based documents. DocRinse may still find private details, but you should review the cleaned result carefully before using AI or sharing the output. Some scanned files may have limited export options.

Can I export a cleaned PDF?

Sometimes. Cleaned PDF export is available only for PDFs that can be regenerated and validated. DOCX, TXT, and pasted text may provide cleaned text or other reviewed export options instead.

Ready to test a sensitive document?

Start with Free Preview, then use the private-alpha workflow when you want to test a real file.