Our methodology

How Albert Checks Facts

A plain-language explanation of how Albert picks claims, weighs sources, and reaches verdicts.

How claims are chosen

Albert monitors what's actually spreading, not what's loudest. Claims reach the queue through three paths:

  • Trending signals. Albert scans current news and social media for claims that are being shared widely right now. Claims people are actively asking about — on the site or in search — are prioritized.
  • User submissions. Anyone can paste a claim into the chat and ask Albert to check it. Frequently asked claims rise in priority.
  • Balance across the political spectrum. Claims are checked evenly whether they originate from the left, the right, or outside politics entirely. Every verdict card records which side of the spectrum the claim skews toward so coverage can be audited.

Albert covers misinformation about disasters and crimes only to correct false narratives — never to share graphic details or identify victims — and waits at least 24 hours on active-casualty events unless an official source has already spoken.

How sources are ranked

Every source Albert cites is assigned one of three quality tiers, based on its domain. A verdict must cite at least two real sources, and at least one of them must be Tier 1 or Tier 2 — a claim can never stand on weak sourcing alone.

Tier 1 — the record of record

Wire services (like AP and Reuters), major news organizations, official government and health agencies, and peer-reviewed scientific journals. When a claim can be checked against a Tier 1 source, that source carries the most weight.

Tier 2 — established regional outlets and fact-checkers

Well-established regional newsrooms and independent fact-checking organizations that follow the International Fact-Checking Network's code of principles. Used to confirm claims when Tier 1 coverage is thin, and always alongside a Tier 1 or Tier 2 source.

Tier 3 — everything else

Blogs, social media accounts, aggregator sites, and any domain we haven't verified. Tier 3 sources may appear for context about what people are saying, but a verdict is never supported by Tier 3 sources alone.

Sources are always listed best-first on a verdict card, so the strongest evidence is the first thing you see.

What each rating means

False. Reliable evidence contradicts the claim — it is not true.
Misleading. The claim contains a grain of truth but presents it in a way that pushes readers to a false conclusion.
Missing Context. The claim is accurate on its face but leaves out key facts you need to understand what it actually means.
Verified True. Multiple reliable sources confirm the claim as stated.
Unverified. There isn't enough reliable evidence yet to either confirm or refute the claim.

What confidence levels mean

High. Several independent, top-tier sources agree, and no credible source disputes the finding.
Medium. The best available sources agree, but independent confirmation is limited or the evidence is still settling.
Low. Evidence is thin, conflicting, or still developing — treat the verdict as a best current reading, not a final word.

Human review

Here is exactly what is and isn't reviewed by a person before it appears on this site:

  • Reviewed before publishing: any verdict that fails Albert's source-quality rules — for example, one that can't find two valid sources or a Tier 1/Tier 2 source — is held back and hidden from the public feed until a human approves or rejects it.
  • Published automatically: verdicts that pass the source-quality rules go live without a human reading them first. They are corrected publicly if an error is found later (see below).
  • Always human-approved: nothing Albert drafts for social media is ever posted without a person approving it first.

Limits and reporting errors

Albert is an AI, and Albert can be wrong. Sources can mislead, evidence can change, and a fast-moving story can look different an hour later. That is why every verdict cites its sources openly, states its confidence, and tells you what evidence would change the verdict.

If you find an error, use the "Report an error" button on any verdict page. Every report goes to a human. When a published verdict's rating or a key fact turns out to be wrong, the change is recorded in the public corrections log — including what the rating used to be.

Editorial independence and funding

Hey Albert is funded by its subscribers. We carry no advertising, accept no sponsorship, and take no money from political parties, campaigns, governments, or the subjects of our fact-checks.

Referral rewards and subscription revenue never influence a rating. Verdicts are reached from sources, not from what anyone pays or prefers, and coverage is kept balanced across the political spectrum. If we get something wrong, we say so on the record.

Dispute a verdict?

Open any verdict page and use the "Report an error" button — every report is read by a human.