De-Identify — Free AI-Powered PHI Masking

Three detection layers work together: regex patterns catch structured data (phone numbers, emails, government IDs, postcodes, dates of birth, street addresses), a neural NER model identifies names, organisations and locations in context, and propagation ensures any value detected once is masked everywhere it appears.

Masked Output

Review the masked output above. PHI tokens (e.g. PHI_a7f3) replace detected names, organisations, locations and other identifiable information. If anything was missed, edit the raw input and re-run, or manually replace in the output before sharing.

What if your AI could do more with this?

CoachIQ takes your de-identified notes and generates pre-session briefs, commitment tracking, score trends, and developmental insights — automatically, session after session.

We'll reach out to set up a personal walkthrough. No spam, no obligation.

How it works

1 Regex Patterns

Structured identifiers — phone numbers, emails, street addresses, government IDs, postcodes, dates of birth, LinkedIn URLs — are caught by purpose-built regular expressions covering AU, US, UK, and Canadian formats.

2 Neural NER

A BERT-based Named Entity Recognition model (running as WASM in your browser) identifies person names, organisations, and locations from natural language context.

3 Propagation

Any value detected by either layer is swept across the full text, ensuring that a name mentioned once at the top is masked everywhere — even in different contexts.

Like what you see?

CoachIQ goes beyond de-identification — it captures your sessions, generates AI coaching reports, and surfaces patterns across your entire practice. ICF-aligned, privacy-first.

Learn More About CoachIQ