PDF Statement Privacy & PII Redaction

PDF statements contain more personal information than CSV files. GigBooks has a multi-layer redaction system to protect your privacy.

Built-in parsers (no AI needed)

For supported banks (Scotiabank, TD, RBC, BMO, CIBC), the built-in parser extracts only the transaction table — it never reads the header area containing your name, account number, or address.

AI extraction (when needed)

If no built-in parser matches your PDF, and you consent to AI extraction, the following process occurs:

Step 1: Text extraction

Raw text is extracted from the PDF.

Step 2: PII redaction

Before the AI sees anything, aggressive redaction replaces:

• Full names → [NAME]
• Account numbers → [ACCT]
• Card numbers → [CARD]
• Addresses → [ADDRESS]
• Phone numbers → [PHONE]
• Postal codes → [POSTAL]

Step 3: AI extraction

The redacted text is sent to Claude AI, which extracts transactions from the remaining merchant names and amounts.

Step 4: Post-extraction scrub

After the AI responds, a second pass removes any remaining digit sequences or name fragments from the transaction descriptions.

Explicit consent required

AI extraction only happens when you explicitly click "Allow AI Extraction". For users who prefer maximum privacy, the built-in parsers and CSV uploads work without any AI involvement.