Redact Text Online

Redact text before you paste it into AI

This free text redaction tool hides personal and secret information before you paste text into ChatGPT, Claude, Gemini, Copilot or any other AI chat. Paste an email, a support ticket, a chat log or a server log, and it finds names, email addresses, phone numbers, street addresses, card numbers, Social Security numbers, API keys, passwords and IP addresses. Each one is swapped for a label like [NAME_1] or [EMAIL_1], so the AI still understands who and what the text is about. When the AI replies, paste its answer back and the real values are put back in. Everything runs in your browser, so nothing you paste is uploaded.

How to redact text

1. Paste your text. Type or paste it into the editor, click Paste, click Upload file or drag a text file onto the editor. Click Sample to see an example. Every item the tool finds is highlighted in the editor in its category’s color.
2. Choose what to hide. Under What to hide, turn categories on or off. Names, emails, phone numbers, addresses, cards and bank details, SSNs and IDs, keys and passwords, IP addresses and usernames are on by default. URLs and dates are off, because they are usually safe to share and the AI often needs them.
3. Hide anything it missed. Select text in the editor and click Hide selected text, or type words under Always hide these words, such as a company name or a project code name. Separate them with commas or new lines, and put a phrase that contains a comma in double quotes.
4. Check and copy. Step 2 shows the redacted text and a list of everything that was hidden. Untick an item to keep it visible, or click Find to see where it is in your text. Then click Copy redacted text and paste it into your AI chat.
5. Put the real values back. Paste the AI’s reply into step 3. Every label, such as [NAME_2], is replaced with the original value, ready to copy.

What the tool finds

• Names: first and last names from a large list of common US and international names, plus names found from context, such as after “Dear”, “Hi” or “Mr.”, in a sign-off like “Best regards, Olivia Bennett”, in “From:” and “To:” lines, in fields like "name": "Emma Lee" and before an email address.
• Emails and phone numbers: any email address, US phone numbers in every common format, such as (415) 555-0132, 415.555.0132 and +1 415 555 0132, and international numbers that start with +.
• Addresses: street addresses with apartment or suite numbers, city, state and ZIP code lines, and PO boxes.
• Cards and bank details: credit and debit card numbers (checked with the Luhn algorithm, so random 16-digit numbers aren’t flagged), CVV codes, expiry dates, IBANs, routing numbers and account numbers.
• SSNs and IDs: Social Security numbers, EINs, dates of birth, and passport, driver’s license, employee, patient and other ID numbers that follow a label.
• Keys and passwords: API keys and tokens from AWS, GitHub, GitLab, Stripe, OpenAI, Anthropic, Google, Slack, Twilio, SendGrid, Hugging Face and many more, JSON Web Tokens, private keys, Authorization headers, cookies, passwords in connection strings like postgres://user:password@host, and values of fields named password, secret, token or api_key in .env, JSON and YAML files.
• IP addresses and usernames: IPv4 and IPv6 addresses, MAC addresses, and usernames in file paths like /Users/emma/ or C:\Users\emma\ and after fields like user=.
• URLs, domains and dates (off by default): web addresses, host names and dates in common formats.

The same value always gets the same label. If “Liam Carter” is [NAME_2] and the text later says just “Liam”, that becomes [NAME_2_FIRST], so the AI can tell they’re the same person.

Labels, fake data or black bars

Under Replace with, choose how hidden values appear:

• Labels, like [EMAIL_1]: the clearest choice for AI chats. The AI knows each label stands for a hidden value, keeps it in its answer, and you can restore the real value afterwards.
• Realistic fake data: swaps each value for a made-up one, such as a different name, an example.com email address, a 555 phone number or a dummy card number like 4000 0000 0000 0001. The text reads naturally, which helps when you share it with people or need the AI to work with realistic-looking data. Fake values can also be restored.
• Black bars: replaces each value with █████, like a redacted document. This can’t be reversed, so use it for text you’re publishing or sending to someone else.

Put the real values back into the AI’s reply

AI chatbots usually keep labels exactly as they are, so a reply like “I’ve drafted a refund email to [NAME_2] at [EMAIL_1]” comes back with the real name and email address filled in. The restore step also handles the small changes AIs sometimes make, such as [NAME 2], NAME_2 without brackets, or \[NAME_2\] in Markdown. If a reply contains a label your key doesn’t have, the tool lists it so you can check it, because the AI may have invented it.

The key that maps labels to real values lives only in your browser tab. To restore a reply later, after you close the tab, click Save key to download it as a JSON file, and click Load key when you need it. The key contains your original data, so keep the file private.

Why redact before using AI chatbots

What you type into an AI chat is sent to the company that runs it. Depending on the service and your settings, it may be stored, reviewed by people or used to train future models. That’s a problem when the text contains a customer’s details, a coworker’s phone number or a production password. Many companies forbid pasting customer data or credentials into AI tools for this reason.

Redacting first lets you still get help: the AI can summarize the email, draft the reply, debug the log or rewrite the document without ever seeing the real names, numbers or keys.

Tips for redacting logs, code and config files

• Server logs: IP addresses, usernames, emails, bearer tokens and session cookies are hidden, while timestamps, error messages and status codes stay readable so the AI can still find the bug.
• Config files and .env files: the values of passwords, secrets, tokens and API keys are hidden, but the key names stay, so the AI can see what’s configured.
• Code: values in code like password = getPassword() or process.env.API_KEY aren’t flagged, because they’re not secrets themselves. A real key pasted into the code is.
• JSON: values inside strings are replaced within their quotes, so the structure stays intact. If a hidden value was a bare number, such as a phone number without quotes, check the result with our JSON Validator before you use it as data.
• Encoded data: secrets can hide inside Base64 text or a JSON Web Token. Decode them first with the Base64 Encoder & Decoder or the JWT Decoder & Encoder to see what’s inside.
• Messy text: if your text has extra spaces, broken lines or odd characters from a PDF, tidy it with the Text Cleaner first.

What automatic redaction can miss

Automatic detection is a strong first pass, not a guarantee. It can miss names it doesn’t recognize, nicknames, unusual ID formats and details that identify someone without being a name or number, such as a job title together with a small town. It can also hide something harmless, such as a word that happens to be a surname. Always read the redacted text before you paste it. Use Hide selected text or Always hide these words for anything it missed, and untick anything that should stay visible.

Is my text private?

Yes. The tool runs entirely in your browser. Your text, the hidden values and the key are never sent to our server, and they aren’t saved in your browser. Close the tab and they’re gone. Only your settings, such as which categories are on and the replacement style, are remembered. Your always-hide words are remembered only if you tick Remember these words on this device.

Frequently asked questions

Is this redaction tool free?

Yes. It’s completely free, with no sign-up and no limit on how much text you redact.

Is it safe to paste personal data into ChatGPT?

It’s safest not to. What you type is sent to OpenAI’s servers, and depending on your settings it may be stored or used to improve their models. Redact names, contact details, account numbers and passwords first, then put the real values back into the reply on your own device.

Will ChatGPT understand text with labels like [NAME_1]?

Yes. AI models handle placeholders well. Because each person or value always gets the same label, the AI can still follow who did what, and it usually repeats the labels in its reply, which makes them easy to restore.

How do I put the real names back into the AI’s answer?

Paste the AI’s reply into step 3, Put the real values back. Every label is replaced with the original value. If you closed the tab, load the key file you saved with Save key first.

Can I redact a file instead of pasting text?

Yes. Click Upload file or drag the file onto the editor. It reads plain-text files up to 20 MB, such as .txt, .log, .csv, .json, .md, .env, .yaml and .eml files. For Word documents and PDFs, copy the text and paste it in.

Does it hide API keys and passwords in code and logs?

Yes. It recognizes the formats of keys from AWS, GitHub, Stripe, OpenAI, Anthropic, Google, Slack and many other services, plus JSON Web Tokens, private keys, bearer tokens, cookies and passwords in connection strings and config files. Long random strings that look like secrets are hidden too.

Can I hide a word that wasn’t detected?

Yes. Select it in the editor and click Hide selected text, or type it under Always hide these words. Every match in the text is then hidden, including matches you paste later.

What’s the difference between redacting and anonymizing?

Redacting removes or masks sensitive details. Anonymizing goes further and makes sure no one can be identified from what’s left, which can mean removing context such as job titles, places and dates. This tool redacts. Read the result and hide any other details that could identify someone.

Is redacted text with black bars reversible?

No. Black bars replace the values completely, and no key is needed or kept for them. Choose Labels or Realistic fake data if you want to restore the real values later.

Does the tool upload or store my text?

No. Everything happens in your browser. Your text and the key exist only in the open tab and disappear when you close it, unless you download the key yourself.