# Tier-one evidence, and why a summary is never a source

Commentaries, firm bulletins and encyclopaedias are useful for finding a rule and useless for proving one. How getax ranks evidence, why a rate can only be validated on the authority's own text, and what that costs.

3 September 2026 · Carol Ireland · 3 min read

![A large navy document with an amber seal, and a faded note card set slightly apart from it.](https://www.getax.io/blog/images/BI-XQZ8bsqfSw_2Aufv)

Most of what is written about tax is written about tax. A firm’s bulletin summarises a budget. An encyclopaedia article condenses a regime into a paragraph. A newsletter reports a threshold change a week after it happened. These documents are how most people first learn that a rule exists, and they are almost never the rule itself.

getax draws the line where a reviewer would. A summary can point at a source. It cannot be one.

## Three tiers, one that counts

Every piece of evidence in the knowledge base carries a tier, decided by who published it.

- **Tier one.** The tax authority’s own pages, the consolidated statute, a treaty as ratified. The text that a tribunal would read.
- **Tier two.** Official but indirect: a ministry circular, a published ruling, an official explanatory note. Authoritative on intent, not always on the number.
- **Tier three.** Everyone else: professional commentary, firm bulletins, encyclopaedias, news. Useful for discovery, and never for proof.

The tier is not a score to be averaged. It is a ceiling. A rate or threshold can be marked validated only when tier-one text supports it. Evidence from a lower tier can at best leave a fact corroborated, however many reviewers agree with it and however confident they sound. The rule binds the arbiter of a review as much as any single reviewer: consensus cannot outrank the evidence.

## Why the number has to come from the authority

A summary is a translation, and translation loses. A bulletin that rounds a rate has dropped the band it applies to. An article that gives last year’s threshold is not wrong, only stale, and nothing on its face says so. A firm that signs a position on the strength of either has signed someone else’s reading of the law.

So the evidence behind a validated fact is the passage itself, kept verbatim and in the original language, with the version of the page it was taken from and the date it was retrieved. When the authority edits the page, the old version is kept and a new one is recorded, so a report delivered last year still resolves to the text it relied on.

## What this costs, and why it is worth it

Insisting on the authority’s own text is slow. Official sites are built for citizens, not for systems; the useful page is often a table inside a portal, a PDF behind a form, or a statute published in a language the search engines rank poorly. getax searches those sites in the country’s own language, because an English query surfaces the commentary and misses the source, and it keeps the exact page it found.

The payoff is in the report. Every material sentence carries a numbered citation, the citation opens to the verbatim passage, and the sources index at the end lists only official documents. A partner reviewing the draft is checking the law, not checking whether somebody summarised the law correctly.

> The question a reviewer asks is not “who said this?” but “where does it say this?” A tier-one citation answers the second question.

The [sample report](/sample-report) shows what that looks like on a fictional engagement: open any citation and read the passage it rests on.