What the Validation Harness Checks Today
The Fake Addresses validation harness runs a fixed set of coherence checks against a sample of freshly generated records on every pull request, and against a larger sample nightly. Checks include: the generated point falls inside its own ZIP Code Tabulation Area; county, state and time zone match that same point by real point-in-polygon lookup; the mailing city matches the USPS last-line city table; the street name exists in the road-network data for that ZCTA; and every reserved-value range — phone, email, SSN, card number — holds with zero exceptions across the sample.
These are exactly the checks published in full, worked example by worked example, on the methodology page, which documents the point-first algorithm the harness verifies rather than restating it here. A build fails outright if any check does not pass, which is what makes the site's coherence claim a tested assertion rather than a description of intended behavior.
Fake Addresses treats this as a genuine differentiator, not an internal detail. A QA engineer choosing between address generators can ask what proof backs a coherence claim, and today the honest answer for Fake Addresses is a nightly-run, fail-closed test suite, documented openly on the methodology page rather than asserted without any supporting evidence at all.
Public Release Status
A public, run-it-yourself release of the Fake Addresses validation harness — usable against this site's own output and against any competitor's, per the original goal of converting a marketing claim into a checkable artifact anyone can rerun themselves — is in progress and not yet published at this URL. Until it ships, the methodology page and the data sources page are the most complete public description of what the harness actually checks and where the data behind each check comes from.
When the public harness ships, this page becomes its home, and the guides section and the glossary will link into it directly rather than only describing it secondhand. Checking back here, or watching the API documentation for a harness-related endpoint, is the fastest way for a QA engineer evaluating Fake Addresses today to know the moment the runnable suite lands rather than just its description.
The same discipline that keeps every generated Fake Addresses record honest applies to this page: it would rather state "not yet published" plainly than describe the harness as available before a visitor can actually run it. That is the same reason the six guides stay unpublished until a named author is confirmed, and it is the same reason the methodology page states a measured, re-checked figure rather than an unconditional guarantee that could quietly turn out to be wrong.
Frequently asked questions
Is the Fake Addresses validation harness open source yet?
Not yet — a public, run-it-yourself release is in progress. Today, the methodology and data sources pages describe in full what the harness checks and how, even though the runnable test suite itself is not yet published at this URL.
What does the validation harness actually check?
Geographic coherence (point-in-polygon county, state and time zone matching), USPS last-line city matching, real street-name presence, format validity against USPS Publication 28, and zero exceptions across every reserved-value range — the full list is on the methodology page.