How Unbunked rates the news

A claim deserves better than a "true" or "false" tossed off in passing. Every article we publish gets a reliability score, built the same way every time. This page explains how it is worked out.

Where we stand

Evidence first: Every criterion rests on sources we opened and cite. The links are there so you can redo the check yourself.
Rigour, not opinion: We assess how sound an article is, never its political side. Orientation is flagged separately and never enters the score: an engaged article can be perfectly reliable.
We own our doubts: When the evidence can't settle it, we say so ("—", unverifiable) rather than hiding the uncertainty behind a made-up number.

The overall score and its criteria

The reliability score runs from 0 to 100. It is the weighted average of six criteria, each rated out of 100, and a fixed formula works it out: two analyses of the same article return the same number. The criteria do not carry equal weight.

Always assessed

Accuracy

30% of the score

Are the article's figures, dates, quotes and claims correct?

Why this criterion: A single false central claim tips the whole article over, which is why this criterion carries more weight than any other.
How we assess it: We isolate the main claims and test them against the sources: quotes not truncated, data not distorted, names and dates consistent. A typo that changes nothing for the reader isn't penalised; a false claim is.

Cross-checking

25% of the score

Are the central facts confirmed by independent external sources?

Why this criterion: A fact repeated a thousand times is still false if it comes from a single origin. We never rely on the article's text alone: we check elsewhere.
How we assess it: Web search and fact-checking databases to find the same fact in at least two reliable, genuinely independent sources, never the same wire copy reprinted.

Cited sources

18% of the score

Are the references the article mentions named, verifiable and independent?

Why this criterion: An article leaning on anonymous or untraceable sources asks for blind trust. We prefer what you can trace back and check yourself.
How we assess it: We examine what the article cites: named sources over anonymous ones, first-hand documents over reprints. A source named in the text counts, even without a hyperlink.

Context

12% of the score

Have essential facts been left out in a way that changes how the article reads?

Why this criterion: You can mislead without lying, by leaving out what's inconvenient. Note: we judge neither tone nor opinion, only omissions that distort understanding.
How we assess it: We check whether the essential context is present, whether a known fact that would change the reading is missing, or whether quotes are taken out of context. No article says everything: a gap that doesn't change the reading costs nothing.

Transparency

10% of the score

Are the author, date and editorial owner identifiable?

Why this criterion: Knowing who is speaking and when is part of the information. An outlet that hides doesn't inspire trust: opacity is itself a flaw.
How we assess it: We look for the author, the publication date and the site's publisher. Disclosed funding is a bonus, never a requirement: almost no outlet publishes it on an article page.

Assessed when relevant

These criteria are only scored when they can be judged reliably; otherwise they are left out rather than producing an arbitrary number.

Recency

5% of the score, when it applies

Is the information up to date rather than stale?

Why this criterion: Information that was accurate yesterday can mislead today. This criterion only applies to time-sensitive topics; it is set aside for timeless content.
How we assess it: We look at the date and check whether more recent facts have made the information obsolete.

Our verdict scale

Every analysis is assigned one of these five verdicts.

The verdict follows from the score: each score range maps to one verdict. The bar for "Reliable" sits high, at 85 out of 100. An article can be serious and well sourced without reaching it, because the label rewards not general seriousness but solidity on the essentials: factual accuracy and corroboration by independent sources.

Reliable85–100 out of 100: Solid, well-sourced and consistent with the facts.
Imprecise60–84 out of 100: Largely true, but incomplete or context-dependent.
Disputable40–59 out of 100: Fragile reliability: major reservations about the facts or sources.
False0–39 out of 100: False or misleading claim.
UnverifiableNo score: Not enough sources to decide.