Content, data & AI methodology

Useful answers grounded in visible sources.

USC uses structured research, public data, automated monitoring and editorial tools to prepare and maintain content. Automation can speed up the work; it does not replace sourcing, uncertainty labels or review rules.

Source hierarchy

USC prioritizes primary and authoritative sources: federal and state agencies, regulators, official statistical releases, product-provider disclosures and contractual partner feeds. Secondary reporting can help identify a topic, but material facts should be checked against the underlying source.

How pages are created

  1. Define the consumer question and intended decision.
  2. Collect source material and structured data.
  3. Use deterministic calculations and templates before generative drafting.
  4. Use automated editorial tools to organize, summarize and connect relevant USC resources.
  5. Validate claims, dates, numbers, links, disclosures and assumptions.
  6. Publish the sources, publication date, modification date and methodology relevant to the page.

Review triggers

Human review is required for legal or regulatory claims, product eligibility, FCRA and consent language, state-specific requirements, safety-sensitive debt guidance, material financial calculations, partner comparisons and conflicting source data.

Update policy

USC updates a canonical page when an existing answer changes. It should not publish a new article merely to repeat the same facts. Rate pages update on the cadence of the underlying source; current-event coverage is published only when the event materially affects credit, borrowing, debt or savings.

Rate and benchmark data

Benchmark rates shown on USC pages come from public sources (the Federal Reserve and the FDIC, retrieved through the St. Louis Fed FRED service). Every figure displays its source and observation date. A value that is older than its freshness limit is hidden or marked unavailable rather than carried forward.

Calculator methodology

Calculators use standard amortization and compounding formulas with the inputs shown. They produce educational estimates, not approvals, quotes, legal advice or individualized financial advice, and they do not model every fee or lender-specific term.

What automation must never do

Automation must not invent rates, approvals, product terms, sources, named experts, consumer outcomes or legal conclusions. Missing information remains missing until it can be supported.

Corrections

Material corrections should be reflected in the page’s update history. Readers can report an issue through the Contact & Complaints page.