Our Methodology

1. How we read your report

When you upload an Equifax, Experian, or TransUnion PDF, the file is parsed in your browser using pdfjs-dist. Only the extracted text — not the original PDF — is sent to our analysis backend. Scanned/image-only PDFs without an OCR text layer cannot be read; we'll tell you when that happens.

2. How the AI evaluates items

Each bureau's text is sent to a large language model with a structured system prompt that asks it to identify accounts and entries that may be:

• Inaccurate (wrong balance, status, dates, or ownership)
• Incomplete (missing required tradeline information)
• Unverifiable (lacking sufficient detail to confirm)
• Outdated under FCRA reporting time limits (15 U.S.C. § 1681c)

The model returns a structured list of items with a short rationale. The model does not access your other reports, your identity, or any external database — only the text you uploaded for that single session.

3. How dispute letters are drafted

For each flagged item you select, CreditKaren drafts a letter addressed to the specific credit bureau, identifying the disputed account and citing the consumer's rights under:

• FCRA § 611 (15 U.S.C. § 1681i) — the bureau's duty to conduct a reasonable investigation, generally within 30 days of a consumer dispute.
• FCRA § 623 (15 U.S.C. § 1681s-2) — the furnisher's duty to investigate disputes referred by a bureau and to correct or delete inaccurate information.
• FCRA § 609 (15 U.S.C. § 1681g) — the consumer's right to request the information in their file.

4. Privacy & data handling

• No account creation. We do not collect your name, address, or SSN to use the analyzer.
• Report text is processed in-memory by the analysis function and is not written to a database.
• Results are cached only in your browser's sessionStorage so a refresh doesn't lose your work; clearing the tab clears the data.
• We use Google Analytics for aggregate, non-PII traffic measurement.

5. Limitations & what we don't do

• We do not guarantee any item will be removed or that your score will change.
• Accurate negative information generally cannot be removed via dispute, even if it's unflattering. The CFPB is explicit about this.
• We are not a credit-repair organization under CROA, a law firm, a lender, or a credit bureau.
• AI-generated analysis can be wrong. Always review flagged items against your own records before sending a dispute letter.

Primary sources

• Fair Credit Reporting Act, 15 U.S.C. § 1681 et seq. (FTC)
• CFPB — How do I dispute an error on my credit report?
• FTC Consumer Advice — Disputing errors on your credit reports
• CFPB Circular 2022-07 — Reasonable investigation of consumer reporting disputes