Text Anonymizer
Paste an interview transcript, focus-group notes or open-ended survey answers and replace the people, places, organizations, emails, phone numbers and dates in it with consistent pseudonyms. Every real value gets the same pseudonym everywhere, and the key that links them is a separate download. It runs in your browser — nothing is uploaded — and it is an aid, not a guarantee: read the output before you share it.
Runs entirely in your browser. Your transcript, your lists and your key file are never uploaded or saved.
Paste the CSV you downloaded before and the same people keep the same pseudonyms across transcripts.
Paste a transcript to anonymize it.
How to Use the Text Anonymizer
- Paste the text. Choose Try a sample to see every feature at once. Nothing leaves the page, and nothing is saved when you close it.
- List what you know. Type the participants' names, the places and the organizations that appear, one per line. Put alternative spellings on one line with a vertical bar —
Priya Nair | Priya | Ms Nair— so they receive one pseudonym. Add=> Aliceif you want to choose the pseudonym yourself. - Switch on the detectors. Emails, phone numbers, web addresses, @handles, IP addresses, postcodes, ID numbers and dates are found by pattern. Speaker labels, titles such as Dr and introductions such as My name is point at names you did not list.
- Review, then export. Check the capitalized words the tool did not replace and add any that identify someone. Copy or download the anonymized text, and download the key file to a different, access-controlled location.
Anonymization, Pseudonymization and What This Tool Does
Replacing a name with Person 1 is pseudonymization: the text no longer names anyone, but whoever holds the key can reverse it. Under the UK and EU GDPR, pseudonymized data is still personal data, so the transcript stays subject to your consent terms and data-management plan. Anonymization goes further: nothing in the text, taken alone or with other information, could single a person out. That usually needs more than swapping names — you generalize or remove details such as a rare job title, a small village or an exact date.
The UK Data Service's guidance, Anonymising data and protecting participants, recommends planning anonymization before you collect data and telling participants at consent what will be done. Direct identifiers — names, addresses, phone numbers, emails — are the easy part, and the part this tool automates. Indirect identifiers are the hard part: a ward manager at a hospital in a named city who started in 2019 may be one person even with the name removed. Only a human who knows the study can judge that.
What the Tool Finds, and What It Cannot
| Kind | How it is found | Known limits |
|---|---|---|
| Emails, web addresses, @handles, IPv4 addresses | Pattern | Obfuscated forms such as name at site dot com are missed |
| Phone numbers | International, US/Canadian, UK-style and 7-digit local formats | Numbers written in words, or in a format outside these, are missed |
| Dates | Day-month-year, month-day-year, month-year, ISO and numeric dates; English month names | Month names in other languages, and phrases like last Tuesday, are not dates to it |
| Postcodes and ID numbers | UK, Canadian and US postcodes; US SSNs, UK National Insurance numbers, card numbers that pass the Luhn check | Other countries' formats |
| Names, places, organizations | Your lists, plus hints: speaker labels, Dr/Ms titles, My name is … | A name nobody listed and nothing hints at is not found |
Always review the output.
The panel Capitalized words that were not replaced lists every capitalized word that appears mid-sentence and was left alone — usually the names and places you forgot. One click adds a word to a list. It cannot find a name that only ever opens a sentence, a name typed in lower case, or a misspelling.
Consistent Pseudonyms and the Key File
Consistency is what keeps the data analysable: if Priya in minute 4 and Priya Nair in minute 40 both become Person 1, a coder can still follow one speaker through the interview. Pseudonyms are numbered in order of first appearance. With Also match first names and surnames on their own on, a surname such as Johnson follows the full name Sarah Johnson — but only when it is capitalized, so a first name like Will does not swallow every will. When two listed people share a first name, the shared part is left unmatched and the tool tells you so.
The key file is a CSV of category, original, pseudonym and number of uses. Paste it back into Key file from an earlier transcript when you anonymize the next interview and the same people keep the same pseudonyms. The key identifies your participants: store it apart from the anonymized text, under the same protections as your raw recordings.
Worked Example
With the sample loaded, Priya Nair: Of course. I'm a ward manager at Northfield Hospital in Leeds. I started on 4 March 2019, after moving up from Sheffield. becomes:
Anonymized
Person 1: Of course. I'm a ward manager at Organisation 1 in Place 1. I started on Date 1, after moving up from Place 2.
The sample's Dr. Okafor, phone number and email address become Person 2, Phone 1 and Email 1 without being listed, and the key file records eight rows. Choosing The year only for dates would leave 2019 — a common generalization, because the year rarely identifies anyone and often matters to the analysis.
Frequently Asked Questions
- Is my transcript uploaded anywhere?
- No. The detection and replacement run in your browser, and the text, your lists and the key file are never sent to a server or stored. Closing or reloading the page clears them, so download the output and the key before you leave.
- What is the difference between anonymization and pseudonymization?
- Pseudonymization replaces identifiers with substitutes and keeps a key that can reverse it, so the data is still personal data under the GDPR. Anonymization means a person can no longer be identified from the data, by anyone, by any reasonably likely means. This tool pseudonymizes; whether the result is anonymous depends on what else the text reveals.
- Can I just run this and publish the transcript?
- No. Pattern-based detection misses identifiers it was not told about, and it cannot judge indirect identifiers such as a rare job in a named place. Read the full output, have a second person check it, and follow your ethics approval and the participants' consent before you share or publish anything.
- How do I keep the same pseudonyms across several interviews?
- Download the key file after the first transcript and paste it into the key box for the next one. People in the key keep their pseudonyms, new people receive the next free number, and the downloaded key grows to cover everyone.
- Why were some names missed?
- Names are found from your lists and from hints in the text. A name nobody listed, that is not a speaker label and carries no title, is invisible to pattern matching. Check the list of capitalized words that were not replaced, and add any that belong.
- Should I replace dates, or keep the year?
- Exact dates, especially birth dates, are strong identifiers; years are weak ones. The year-only option keeps the timeline you may need for analysis. Choose a label instead if even the year could point to one person in a small group.
- What should I do with the key file?
- Treat it as personal data. Keep it away from the anonymized transcripts, restrict who can open it, and delete it when your data-management plan says the link between participants and pseudonyms is no longer needed.