Editorial Policy

Workflow

Calculator Nova content is AI-assisted. There is no human reviewer; instead, an automated pipeline of quality gates stands between a page and a live deploy. We do not claim a human review that does not occur.

The eight gates that replace the human reviewer

1. Unit tests — every formula module has passing unit tests, each including an externally-sourced reference value and edge cases.
2. Citation validator — every cited .gov/.edu/standards URL must resolve and match its source; hallucinated or dead citations block the build.
3. Structured-data validator — JSON-LD is checked with the Google Rich Results Test API.
4. Link checker — no broken internal or external links.
5. Freshness check — per-data-type freshness windows (currency 24h, fuel weekly, CPI monthly, tax annual).
6. Accessibility gate — axe-core, zero critical violations.
7. Performance budget — LCP < 2.5s, CLS < 0.1, INP < 200ms, JS < 50KB/page.
8. Visual regression — Playwright pixel-diff on chart primitives.

A failing gate blocks merge with no override.

Sources

Citations are limited to real, resolving .gov, .edu, and standards-body sources. No fabricated “per IRS Pub X” or “per ASTM Y” claims.

No fabricated authors

We do not invent credentials or persons. The team byline is used unless a real named contributor is involved.

Content boundaries

There is no health or medical content. Sports tools compute numbers only and never prescribe training (“you should train at X”). Finance tools carry a “not financial advice” disclaimer.