Practical guide · By Aurelio Avila · September 6, 2026

Redact sensitive data before ChatGPT.
Keep the question.

You need help rewriting a client email or debugging an error. The AI needs the problem and enough context to understand it. It usually does not need a real email address, a live credential or a customer's account number.

Short on time? Check a prompt on the web for free, or read how Redaxa removes personal data and secrets before the text goes anywhere.

1. Reduce what you share first

Copy the relevant paragraph or error message, not the full mailbox, customer record or configuration file. Replace names and internal project references with consistent labels. Keep the sequence of events and the technical details that matter to the question.

Check your organization's rules before sharing work material with any external service, including a scanning service.

2. Check more than names

Credentials: API keys, passwords, authorization headers, tokens and private keys.
Identifiers: email addresses, phone numbers, client names and internal project terms.
Financial details: payment card numbers, IBANs and billing references.
Context: combinations of details that identify someone even after their name is removed.

A scanner can help find patterns. It cannot decide whether you have permission to share a document, or reliably recognize every confidential fact.

3. Preserve useful context

This example is fictional and manually redacted; it is not a live scan or a claim that every field is automatically detected.

Before

Draft a reply to alex@example.com about project Cedar.
The staging request fails with HTTP 401.
The config contains password=demo-only-not-a-real-secret.
Ask for the deployment time and request ID.

After review

Draft a reply to [CLIENT_EMAIL] about [PROJECT].
The staging request fails with HTTP 401.
The config contains password=[REDACTED].
Ask for the deployment time and request ID.

The task, error and next steps remain. The recipient, project identifier and credential do not. Add project names to protected terms when using Redaxa; do not assume generic detection knows your internal vocabulary.

4. Review before sending

Read the cleaned version yourself. Check quoted replies, signatures and surrounding context. No findings means that the enabled checks found nothing; it is not a guarantee that the text is safe.

If you already exposed a real credential, redacting a later copy does not invalidate it. Use the provider's credential controls to revoke or rotate it.

Where Redaxa fits

Redaxa by Aurelio Avila checks supported patterns in prompts and offers a redacted version. The Chrome extension works inside supported AI sites; a web scanner and Windows app are also available.

Where the check runs: free checks in the Chrome extension run inside your browser, and the text is not sent anywhere. The web scanner, the Windows app and the extension with an active plan send prompt text to Redaxa's backend for scanning; it is not forwarded to a third-party AI service and is not stored. Read the privacy policy before using it with sensitive work.

View the Chrome extension

The extension's free checks need no account and have no daily limit. Protected terms and team policies need Pro or Business; Pro and Business start with a 3-day free trial: a card or another payment method is needed, nothing is charged for 3 days, and you can cancel before it ends from Manage subscription. See pricing.

Redact sensitive data before ChatGPT.Keep the question.