Use cases · Healthcare
Anonymize patient reports before using an AI
Anonymize any document before it reaches an AI. Nothing leaves your device. Discharge summaries, referral letters and notes keep their clinical content; the identifiers become labels.
- 0 network requests
- Works with Wi‑Fi off
- Free, no account
- Restore names afterwards
Why it matters
Why clinicians anonymize first
Health data is a special category
GDPR and KVKK treat health information as sensitive. A summary request should not carry a patient's name, ID and date of birth to an external service.
The clinical content is what matters
Symptoms, findings, medication and plan are what you want summarized or translated. The patient can be Patient 1.
Identifiers hide in headers and footers
Hospital letterheads, protocol numbers and physician names sit in headers, footers and tables. All of them are scanned in DOCX and PPTX.
You restore the identity locally
The mapping table stays on your device. Restore the names in the answer with the de-anonymizer when the text goes back into the record.
How to do it
Four steps, all in your browser
- 1
Upload the file or paste the text
Word, PDF, PowerPoint, Excel, CSV, JSON, HTML, Markdown or plain text. Up to 10 files, 50 MB each. The file is opened in your browser tab and never sent anywhere.
- 2
Choose the rule
Decide how each category is rewritten: labels like Person 1 / Company A, realistic fake values, masking or redaction. Optionally enable smart detection and add words to your dictionary.
- 3
Review what will change
Names, companies, emails, phone numbers, IDs and IBANs are listed with their replacements. Turn items off, add missed ones, correct a category. Nothing changes until you confirm.
- 4
Download and paste into the AI
You get the same file back, a Markdown version to paste into the chat, and a mapping table. Paste the AI's answer into the de-anonymizer to put the original names back.
What to replace
Checklist for this domain
The tool detects most of these automatically. Add the rest to your dictionary so they are replaced consistently.
- Patient names and initials
- Dates of birth and admission dates
- National ID and insurance numbers
- Protocol, file and sample numbers
- Hospital, clinic and department names
- Physician and staff names
- Addresses and phone numbers
- Relatives' names in the history
- Rare conditions combined with locations, if identifying
Before and after
What the AI sees
Original
Ali Yılmaz (TC 10000000146, born 03.02.1958) was admitted to Acıbadem Maslak on 12.01.2026 with chest pain. Dr. Zeynep Kurt ordered an ECG.
Anonymized
Person 1 (TC ID 1, born Date 1) was admitted to Organization A on Date 2 with chest pain. Dr. Person 2 ordered an ECG.
Try it now
Anonymize your document here
Nothing is uploaded. Switch Wi‑Fi off if you want to check.
- 1Upload
- 2Choose rule
- 3Review
- 4Download
Upload
Drop files here or choose
Up to 10 files · 50 MB each · files stay on your device
- docx
- pptx
- xlsx
- txt
- md
- csv
- json
- html
Old .doc, .ppt and .xls files: save them as .docx, .pptx or .xlsx first.
Questions
Common questions
Are national ID numbers validated?
Yes. TC Kimlik No and similar numbers are checked with their checksum, which keeps false positives low.
Can I keep the dates but remove the names?
Yes. Categories are toggled individually. Keep dates on if the timeline matters and off if it identifies the patient.
What about scanned reports?
Scanned PDFs without a text layer are not supported. Run OCR first, then anonymize the text or DOCX.
Does the report leave the device?
No. Everything runs in your browser; 0 network requests during processing.
Other use cases