Emotional framing is not evidence, and it's not disqualifying either — a claim can be upsetting and true, or calm and false. What decides whether something gets published here is whether it survives being traced back to where it came from. That's the whole standard. Everything else on this page is what "traced back" means in practice.
The six questions
Where does this actually come from?
Not "a study found" — which study, which authors, which year, published where. If a claim can't be traced to a specific, named, citable source, it doesn't get treated as fact here, no matter how plausible it sounds or how many accounts are repeating it.
What did the source actually measure?
A study measures something specific — a correlation, a within-family comparison, a retweet rate. The plain-language summary of that finding is often looser than the finding itself. This page reads the paper, not just the press release or the caption.
Where does the source's claim end and the narrator's argument begin?
This is the seam most fact-checking skips. A video or post will often cite a real, correctly-described study — and then quietly extend its conclusion to cover something the study never tested. Both halves can be true independently: the study is real, and the extension is the storyteller's own argument. Case 001 on this site is built entirely around one example of this.
Has anyone failed to replicate it?
Single studies, even good ones, aren't final answers. Where a widely-cited finding has a documented replication attempt — successful or not — that gets reported alongside the original, not buried. See the motivated-numeracy study page for an example where a large replication came back negative, and both results are shown.
What would change this conclusion?
If a claim is structured so that no possible evidence could ever count against it, that's a signal on its own — regardless of which side it favors. Falsifiable claims get a fair hearing even when they're unwelcome. Unfalsifiable ones get named as such.
What does the other side's strongest version look like?
Not the weakest opposing argument — the strongest one, stated the way its own defenders would state it. A correction that only ever rebuts the worst version of a disagreement isn't a correction, it's a strawman with a citation attached.
What "holds," "caveat," and "overreach" mean
Every exhibit on a case page carries one of three tags. They describe the claim, never the person or channel that made it — the same source can produce a claim that holds and a claim that overreaches in the same piece of content.
The claim matches what the cited source actually says, within normal rounding and plain-language paraphrase.
The core fact is accurate, but something material is missing — a contested replication, a narrower scope than implied, a name or number transcribed wrong.
The source is real, but the claim built on top of it goes further than the source supports — a correlation presented as the source's own causal claim, for instance.
No underlying event or document exists at all — checked against the primary record (hearing footage, court filings, official transcripts) and not found. Different from "overreach": there's no real source being stretched, because there's no real source.
When we're wrong
Corrections get made in place, on the page, with a visible note — not quietly edited out. A publication that never has to correct anything usually isn't checking hard enough, not the other way around. If a source, a number, or a framing on this site turns out to be wrong, that's not a reason to take the page down; it's the reason the page exists.
Labels like "biased" or "propaganda" are often used to end a conversation before the evidence gets examined — on every side of every argument. The fix isn't to argue louder. It's to make the sourcing checkable by anyone, including someone who starts out skeptical of the person making the claim. That's the actual goal of this whole page.