How we assess a claim, and what our grades mean.
The grade scale
Every debunk carries one grade. The letter is a shorthand; the word is the verdict.
| Grade | Meaning |
|---|---|
| A · True | The claim is accurate and well-supported by the evidence. |
| B · Mostly True | Broadly accurate, with caveats that matter but do not overturn it. |
| C · Mixed / Unproven | The evidence is genuinely split, or there is not yet enough of it to say. |
| D · Misleading | Technically defensible, but framed to produce a false impression. |
| E · False | Contradicted by the evidence. |
Two of these deserve emphasis, because they are the ones most often skipped elsewhere.
C is not a cop-out. "We do not know yet" is a real and common state of the evidence, and reporting it honestly is more useful than manufacturing a verdict. A claim graded C is one where the honest answer is that the science has not settled.
D is not a softer E. Misleading claims are usually built from true components — a real study, a real statistic, a real mechanism — arranged to imply something the evidence does not support. These are frequently more persuasive, and therefore more consequential, than claims that are simply false.
The evidence hierarchy
Weighted roughly in this order:
- Systematic reviews and meta-analyses of randomised controlled trials.
- Individual randomised controlled trials.
- Large prospective cohort and observational studies.
- Position statements from public-health bodies and professional associations.
- Mechanistic, animal, and in-vitro studies — which describe what could happen, not what does.
- Case reports and clinical anecdote.
Nothing below the line: personal testimony, influencer experience, and "I felt better" are not evidence of a general effect, however sincerely offered.
A recurring pattern worth naming: a great many viral claims rest on level 5 evidence presented as though it were level 1. A compound that does something interesting to cells in a dish has not been shown to do that thing in a person.
How we source
- We link the original claim wherever it was actually made, so you can see it in context rather than take our summary of it.
- We link and archive our sources, preferring the primary study over any reporting about it. If we cite a finding, we have read the source, not just coverage of the source.
- Sources are archived because links rot and pages get quietly edited. The archived copy is the durable record of what a page said when we assessed it.
- Research tools — including AI research tools — are used to find candidate claims and sources. Nothing they summarise reaches publication unverified against the primary source. The standard we apply to influencer claims applies to our own inputs.
What we will not do
We will not grade a claim we cannot source. If the evidence is not there, the piece does not run — an empty verdict is worse than no verdict.