Anonymise data before you feed it to an AI model
This page is preset to the Fake values style: instead of bracketed tokens, each real value is replaced with a plausible stand-in — “Jordan Blake”, “person3@example.com”, “(555) 555-0102”. The fake phone numbers come from the 555-0100 to 555-0199 block that the North American Numbering Plan reserves for fictitious use, so no stand-in ever rings a real person.
Fake values are the right choice when you want the model to reason about the text as prose: drafting replies, summarising a transcript, classifying tickets. Numbered tokens are better when you want to be certain, at a glance, that nothing real is left.
Your text
Safe to paste
Preset: realistic stand-in values instead of tokens, so model output reads naturally.
Mapping — 0 values (only visible in this tab)
| Placeholder | Real value | Uses |
|---|
Check a few rows before you paste — this is the fastest way to spot a detector that grabbed the wrong thing. Nothing here is stored or sent anywhere.
What to look for — 30 detectors, custom terms, exceptions
People & organisations
Government & health IDs
Financial
Technical & secrets
Location
Step 3 — paste the AI’s answer here to put the real values back
The model replies using the placeholders. Paste its answer below and every
[NAME_1] becomes the real name again — using the mapping held in this tab, which
is the only copy that exists.
How it works
- Open this page — the detectors are already set for “Anonymise data for AI”.
- Paste your text, or drop a .txt, .md, .csv, .json or .log file onto the panel.
- Check the highlights: every match is coloured by type, and one click keeps a type as-is.
- Copy the safe version and paste it into the AI tool of your choice.
- Paste the model’s reply back into step 3 to restore the real values.
Consistency is what makes fake data usable
A stand-in is only useful if it is stable. Every occurrence of the same email address becomes the same fake address, in this text and for as long as the tab stays open, so a model can still tell that rows 1 and 4 are different customers and that row 1 mentions the same person twice. Randomising each occurrence would destroy exactly the structure you are asking the model to analyse.
The mapping is deterministic in the other direction too: the same input with the same settings produces the same output. That matters if you are pasting several batches over an afternoon and want the sixth batch to talk about the same “Jordan Blake” as the first — keep the tab open and it will.
Where fake values are the wrong tool
If you are publishing a dataset, handing it to a third party, or relying on a legal definition of anonymised data, this is not enough. Substitution with a stored mapping is pseudonymisation. Under GDPR Recital 26 and the EDPB’s Guidelines 01/2025, pseudonymised data stays personal data while the additional information that reverses it exists — even if the only copy is a browser tab. The EDPB even names the environment holding the mapping a “pseudonymisation domain” and expects it to be protected accordingly.
Real anonymisation also has to survive the mosaic effect: a scrubbed record can still be identifying because of what remains. “The 61-year-old orthopaedic surgeon in our Cheltenham branch who resigned in April” contains no direct identifier and identifies exactly one person. No pattern matcher can catch that. Read what is left, not just what was removed.
Tips
- Analysing a spreadsheet? Paste one column at a time — placeholders stay stable across pastes as long as the tab is open.
- Add the company’s own name to custom terms; brand names carry more identifying weight in a small dataset than most people expect.
- For anything you will publish, drop the free-text fields entirely rather than scrubbing them — narrative text is where re-identification usually starts.
Detectors are pattern-based and imperfect: read the highlighted output before you paste it anywhere, and confirm your own confidentiality obligations against the current rules for your profession, employer or jurisdiction. Compiled 2026-07-29.
Frequently asked questions
Are the fake names ever real people?
The pool is invented first names and surnames combined deliberately (“Quinn Delacroix”, “Rowan Petrov”), and the emails all use example.com, which IANA reserves for documentation. Any resemblance to a real person is coincidence, and no stand-in address or number can reach anyone.
Can I undo it later?
Yes, while the tab is open — paste the model’s answer into step 3 and the real values come back. If you need to restore after closing the tab, save the key file first. That file is the mapping in plain text: treat it like a password and delete it when you are done.
Does this make my data GDPR-compliant?
It is a technical measure, not a compliance verdict. Pseudonymisation is explicitly recognised in the GDPR as a safeguard, but pseudonymised data remains personal data, so your lawful basis, retention and transfer obligations still apply. Anonymised data falls outside the GDPR — but only if re-identification is not reasonably possible for anyone, which a reversible mapping does not satisfy.