Methodology

Last updated: 2026.

This page is the canonical statement of how we work. Anyone can read it, anyone can hold us to it.

What we cover

We focus our coverage on factual claims made by Donald Trump, members of his administration, and prominent figures aligned with the MAGA political movement. We chose this focus because these figures hold or seek significant political power, the volume of disputed factual claims from this cohort is high, and existing generalist fact-checkers cannot give every claim the time it deserves.

We do not fact-check matters of opinion, taste, or personal belief. We fact-check statements that purport to describe reality.

What we rate

Every fact-check on this site rates a single, specific factual claim. We do not rate the speaker, the speech, or the political program. We rate the claim.

Our rating scale

• True. Supported by primary, verifiable sources. No material qualifier missing.
• Mostly True. Essentially correct, but needs clarification or context that changes its impact.
• Half True. Partially accurate, but leaves out important context or combines true and false elements.
• Mostly False. Contains a kernel of truth but is misleading on the central point.
• False. Contradicts primary, verifiable sources.
• Pants on Fire. Not only false but absurd or showing clear signs of fabrication.
• Unverifiable. Cannot currently be verified. Sources are unavailable or contradictory.

A note on "lie." We do not use the word lie in our ratings or headlines. "Lie" implies intent — a state of mind we cannot observe and cannot prove. We say a claim is false when it contradicts the evidence. Whether the speaker knew it was false is a question we leave to the reader.

How we source

Every fact-check is built from primary sources wherever they exist: government data, official transcripts, court records, original publications, direct video or audio, original documents. Secondary sources are starting points, not destinations. Every numerical or factual assertion in our fact-check traces back to a primary source we link to.

Our process

1. Capture. A claim is identified from a transcript, post, or article. Exact quote, speaker, date, venue, and source link are recorded.
2. Extraction. The factual claim is isolated from rhetoric and split into atomic claims if compound.
3. Research. Primary sources are gathered. We seek both confirming and disconfirming evidence before reaching a rating.
4. Drafting. A draft fact-check is written. Drafts on this site are produced with assistance from large language models; every draft is reviewed by a human editor.
5. Editorial review. A human editor verifies that every cited source actually says what the draft claims. The editor revises, downgrades, upgrades, or rejects the rating.
6. Publication. The fact-check is published with a permanent URL, the source list, the original quote in context, and the date.
7. Corrections. Errors are corrected visibly at the top of the page with the date. We never silently edit a rating.

Our use of AI

Drafts on this site are produced with the help of large language models (currently Anthropic's Claude). Specifically: the model receives the claim and source documents we have already gathered, and produces a draft summary, a tentative rating, and citations. A human editor then independently verifies every cited source before publication. The model is never the sole judge of a claim's truth.

Conflicts of interest

We do not accept funding from political campaigns, candidates, party committees, PACs, or 501(c)(4)s aligned with any candidate or party. Funding sources are disclosed on the Funding page.

Corrections and complaints

If you believe a fact-check on this site is wrong, email corrections@magashmaga.com with a link to the fact-check, the specific assertion you believe is incorrect, and supporting evidence. We respond within 7 days. Corrections, when warranted, are published at the top of the affected page.

What we will not do

• We will not rate a person. We rate claims.
• We will not call a claim a lie. We rate claims based on evidence; we do not infer state of mind.
• We will not rate a claim based on the political identity of the speaker. The same claim from any speaker would receive the same rating.
• We will not publish a fact-check our editor has not personally verified.