Source methodology
CertiWatt is designed to answer a narrow question: whether a specific power-bank model can be carried on a specific air-travel route under the cited public rules currently tracked by the system.
The engine first tries to match the traveler input to a catalog model with recorded capacity, watt-hours, voltage basis, certifications, manufacturer evidence, and known aliases. If the model cannot be resolved, the verdict can remain insufficient.
Watt-hours are treated as the aviation decision unit. When a manufacturer publishes watt-hours, that value is preferred. When only milliamp-hours and voltage are available, the catalog records the watt-hour basis so downstream verdicts can explain the calculation.
The ruleset evaluates baseline lithium-battery constraints, departure-country overlays, airline policy, quantity limits, carry-on handling, approval requirements, label requirements, and local certification rules where active.
Active recall records are scoped to exact model IDs and, where available, serial or batch evidence. Recall impact can override an otherwise capacity-compliant answer because safety status is a separate condition from watt-hours.
The output is allowed, conditional, banned, or insufficient. When multiple active constraints apply, the stricter supported result wins and the response retains the evidence needed to audit that path.
Verdicts expose citations for the regulator, airline, manufacturer, or recall sources used in the decision. Public pages and agent surfaces also expose citation kind, jurisdiction, and pulled-at metadata where available.
Regulator and aviation authority pages are preferred for jurisdictional constraints. Airline policy pages are used for carrier-specific passenger rules.
Official product pages, manuals, support articles, and recall pages are preferred for model identity, capacity, certification, and affected serial evidence.
News or third-party reporting can explain context but should not replace official policy, specification, or recall evidence when those sources are available.
Source freshness is monitored through the ingestion workflow. The public scorecard reports when sources were checked, whether content was verified, and whether a source is stale, changed, unavailable, or failing.
Golden tests and contract tests are used to detect false allowed, false banned, citation, recall, and public API regressions before release.