Methodology

This is not a padded legal disclaimer — it is a checkable account of how every verdict on this site is produced, and of when we deliberately refuse to produce one. If you're making a travel or shipping decision from this data, read this page first.

1. Two unrelated switches drive every verdict

Most “can I bring / can I ship” questions get answered wrong because the respondent only sees one switch:

• Switch A — destination import law. Enforced by customs, biosecurity, and drug authorities. It answers: does this country let it in?
• Switch B — carrier acceptance policy. Enforced by airlines, posts, and couriers. It answers: does this route take it today?

B is usually stricter than A (airlines on lithium batteries are the classic case), and a B-answer guarantees nothing about A. This site models A as 105 rules and B as 12 carrier-acceptance entries, answers both on one page, and marks explicitly where a carrier's answer does not cover import law.

2. Items are modeled as attributes, not as a list of conclusions

Each of 44 item profiles only describes what the thing is: origin, source detail, processing, storage, battery energy, blade length, controlled substances, and so on. Rules describe what verdict items with given attributes receive in a given country and channel. A verdict is the solved result of attributes against rules.

One reason only: conclusions must re-derive when the item changes. “Jerky: no” and “commercially sealed beef product: maybe yes” are not two isolated rumors — they are the same attribute set solved twice with different values. That is also the technical precondition for the paid dossier's unlock paths: change one attribute, re-solve, instead of recalling what some forum said.

3. When multiple rules match, the strictest one governs

The severity order (strictest first). Aggregation always resolves toward the strict side, because a lenient rule overriding a strict one can mean destruction or criminal liability, while the reverse only costs you a step:

• Prohibited(rank 50)
• Self-verification required(rank 40)
• Prior permit required(rank 30)
• Declaration required(rank 20)
• Conditional(rank 10)
• Generally permitted(rank 0)

“Verify for yourself” sits deliberately between “prior permission” and “prohibited”. In compliance, not-knowing is more dangerous than knowing-you-need-a-permit — the latter can be planned, the former cannot be evidenced. But it must never be dressed up as an explicit prohibition — that would be fabrication.

4. Freshness: data turns conservative on its own

Every rule carries verified_at (the date a human checked it against the source); every citation carries cadence_days (the required re-check window). The engine computes confidence from these and enforces one hard constraint: only source-verified data may support a lenient verdict.

• Verified against source：This verdict was manually compared against the official text, with the verification date recorded.
• Source due for re-check：The rule itself is verified, but the cited source is past its recommended re-check window — re-confirm before a trip that matters.
• Verification expired：The last verification exceeded the re-check cadence this source requires; this page answers on the conservative side.
• Not yet verified：This entry has not been verified against the official text, so it is never used as grounds for release; we assume the strictest reading.

The consequence: any “allowed / generally permitted” answer you get here comes from source text we verified inside its re-check window. When “verify for yourself” appears, we aren't fobbing you off — we're telling you this data point isn't decision-grade yet.

Current state: 105 of 105 rules are source-verified; the rest keep demoting. 82 cited documents: 11 pieces of primary law, 63 regulator publications, 6 carrier terms. 10 unlock paths — each is re-solved and compared against its declared expected outcome; any drift enters the review queue.

5. Language models produce no verdicts

This is the product's boundary, not a slogan. The engine emits structured data; models are allowed only to rewrite existing conclusions into better prose — never to produce, alter, or infer a determination. In this category one hallucination isn't an inaccurate answer, it's a destroyed box of goods or an on-the-spot fine at the border.

6. What we do not do

• We don't host or mirror official texts — we link them and mark the re-check cadence. The source remains the authority.
• We don't give legal advice and don't speak for any regulator or carrier.
• We don't cover controlled drugs, endangered species (CITES), or commercial-volume imports of weapons and ammunition — go straight to a licensed professional for those.
• We don't fill data gaps to make a page look complete.

7. How corrections work

If you flag a wrong verdict with an official link, and we confirm it, we will: correct the data, publish a dated correction note on the page, and re-review every combination the same rule touches. Submit through the contact page with the institution name and link.

This page describes BorderPrep's own data and engineering constraints and involves no regulator. Rules and verification progress also appear in the liability section referenced by the terms.