← Blog  ·  April 30, 2026  ·  6 min read

How to remove personal information from text

Personal information (or PII — Personally Identifiable Information) is any data that can be used to identify an individual: their name, email address, phone number, home address, Social Security number, date of birth, and more. Removing it from text before sharing, storing, or processing it is a core requirement of privacy regulations like GDPR, CCPA, and HIPAA.

This guide covers the three main approaches: the free online tool (paste and go), the Python API (automated pipelines), and manual regex (when you need full control).

What counts as personal information in text?

Note: generic role mentions ("the doctor", "the manager"), company names, and public figures in their public capacity are not personal information and should not be redacted.

Method 1: Free online tool (fastest, no code)

If you have a single document, email, support ticket, or any text you want to clean up, the free online tool at PrivacyFilter.run is the fastest option.

Example

Input:

Hi, I'm Sarah Johnson (sarah.johnson@acme.com, +1 555-842-1234).
My address is 14 Elm Street, Boston MA 02118 and my SSN is 543-21-6789.

Output (Replace mode):

Hi, I'm [PERSON_1] ([EMAIL_2], [PHONE_3]).
My address is [ADDRESS_4] and my SSN is [SSN_5].

All five pieces of personal information detected and replaced in one click. The original text was never stored on any server.

Method 2: Python API (automated pipelines)

For removing personal information at scale — thousands of support tickets, a log export, a CSV of customer data — the PrivacyFilter REST API handles the same detection logic with a single HTTP call:

import httpx

LICENSE_KEY = "your-license-key"  # get at privacyfilter.run

def remove_pii(text: str) -> str:
    response = httpx.post(
        "https://privacyfilter.run/api/redact",
        json={
            "text": text,
            "license_key": LICENSE_KEY,
            "mode": "replace"  # or "mask" or "tag"
        },
        timeout=15,
    )
    response.raise_for_status()
    return response.json()["redacted_text"]

# Remove PII from a single string
clean_text = remove_pii("Call Jane Doe at 555-0192 to confirm order #A1042.")
print(clean_text)
# → "Call [PERSON_1] at [PHONE_2] to confirm order #A1042."

Batch processing (multiple documents)

import httpx

documents = [
    {"id": "doc_1", "text": "Customer Alice Walker emailed alice@shop.io..."},
    {"id": "doc_2", "text": "Invoice for Robert Chen, 22 Baker St, London..."},
    # up to 20 documents per batch call
]

response = httpx.post(
    "https://privacyfilter.run/api/redact/batch",
    json={"documents": documents, "license_key": LICENSE_KEY, "mode": "replace"},
    timeout=60,
).raise_for_status().json()

for result in response["results"]:
    print(result["id"], "→", result["redacted_text"][:60])

Method 3: Manual regex (when you control the format)

Regex is the right choice only when your text has a predictable, well-known format — such as structured CSV exports where emails always appear in column 3. It fails completely for unstructured text like emails or support tickets, where "Mary from accounts" is a name that no regex can reliably catch.

import re

EMAIL_RE  = re.compile(r'\b[A-Za-z0-9._%+\-]+@[A-Za-z0-9.\-]+\.[A-Za-z]{2,}\b')
PHONE_RE  = re.compile(r'\+?1?\s?[\(]?\d{3}[\)]?[\s\-\.]?\d{3}[\s\-\.]?\d{4}')
SSN_RE    = re.compile(r'\b\d{3}[-\s]?\d{2}[-\s]?\d{4}\b')

def regex_redact(text: str) -> str:
    text = EMAIL_RE.sub('[EMAIL]', text)
    text = PHONE_RE.sub('[PHONE]', text)
    text = SSN_RE.sub('[SSN]', text)
    return text

When you need to remove PII permanently vs. pseudonymize

There are two different techniques depending on your use case:

• Anonymization — personal data is removed or replaced with random values (e.g., ████). Irreversible. Required for GDPR "no-longer-personal-data" status.
• Pseudonymization — personal data is replaced with consistent placeholders (e.g., [PERSON_1]). Reversible if you keep the mapping. Still considered personal data under GDPR but reduces risk. Useful when you need to re-insert names later (e.g., personalizing an LLM response).

PrivacyFilter supports both: mode=mask for anonymization, mode=replace for pseudonymization (the entity table gives you the mapping to reverse it).