Prompt Anonymizer — Replace Names & Entities

Replace person names, organization names, product codes with consistent [PERSON_1], [ORG_1] placeholders so the LLM sees structure but not specifics. Useful when the structure of an internal scenario matters but the actual names shouldn't go to an external API.

How to use the Prompt Anonymizer — Replace Names & Entities

Paste the text. List the entities to replace, optionally with a type prefix (person:, org:, product:). Each unique entity gets a numbered placeholder. Mapping is preserved so you can de-anonymize the LLM's response by reversing the substitution.

Replacing names before sending to an API

Sometimes the structure of an internal scenario is what you want an LLM to reason about, but the real names — people, companies, product codes — shouldn't leave your network. Anonymizing replaces each entity with a consistent placeholder like [PERSON_1] or [ORG_1], so the model sees the relationships and roles without the identifying specifics.

Because the mapping is consistent and reversible, the same name always maps to the same placeholder, and you can swap the originals back into the model's response afterward. This is a deliberate, entity-list approach: you say what to replace. For pattern-based scrubbing of emails, phone numbers, and card numbers, use the prompt PII redactor instead.

Common use cases

• Internal scenarios — reason over a real situation without sending real names to an API.
• Consistent placeholders — keep one entity mapped to one token across a long document.
• Reversible workflows — de-anonymize the model's reply by reversing the mapping.
• Sharing examples — strip identifying names before pasting a case into a prompt.
• Vendor and product codes — mask internal identifiers that shouldn't be exposed.