Facts receive scores.
A checkable claim describes something that could be supported or contradicted by evidence: an event, quotation, measurement, date, amount, cause, or documented action.
How a claim
becomes a score.
FakeNewsZero.com measures how strongly the available evidence supports a factual claim. The score is a structured summary of evidence—not the probability that a statement is true.
Every result follows the same path: decide what can be checked, break it into parts, examine the evidence, combine the scores, and apply limits where confidence would otherwise be overstated.
Score at a glance
Turn the number into a readable conclusion.
90–100Verified
Strong, direct evidence that satisfies the claim-specific standard.
75–89Highly Credible
Good evidence with limited remaining uncertainty.
55–74Credible but Unconfirmed
Some support, but an important gap remains.
35–54Insufficient Evidence
The available evidence is too limited or mixed.
15–34Likely Misleading
Key parts are unsupported, distorted, or contradicted.
0–14False / Contradicted
Reliable evidence directly conflicts with the claim.
01
First, decide what can be checked.
Opinions do not.
Personal judgments such as “this person is unintelligent” are not scored as true or false. If a statement mixes opinion with factual premises, only the factual parts are checked.
Vague attribution needs clarification.
For phrases such as “someone said,” the user chooses whether to check the words themselves or whether a named person or source actually said them.
02
Break the statement into atomic claims.
A sentence can contain several independently testable ideas. The system separates the core assertion from dates, quantities, causes, locations, groups, and attributions. Each part must be able to stand or fail on its own.
Illustrative five-part decomposition
“A 2025 Stanford study found that high-school students who slept at least eight hours scored 20% higher on math tests because sleep improved memory.”
- Source
Did Stanford publish or conduct the cited study?
- Timing
Was the study published in 2025?
- Population
Did the study examine high-school students?
- Measured result
Did students sleeping at least eight hours score 20% higher on math tests?
- Causal explanation
Does the evidence show that improved memory caused the score difference?
03
Give each part a score and a weight.
Choose the evidence standard.
Each part is classified as a stable reference fact, authoritative self-record, current event, contested or high-stakes claim, numerical claim, or something that cannot presently be verified. The claim type determines whether one source of record is sufficient or independent confirmation is required.
Evidence score and weight.
Each atomic claim receives a score from 0 to 100 based on authority, directness, independence, recency, context, and contradiction. Its percentage weight represents how much that part matters to the full statement, and all weights total 100%.
A small date error should not erase a well-supported central claim. A failed core claim should matter much more than a minor detail.
The calculation
Adjust each part before combining it.
Applies the evidence limit required by that claim type. A verified source-of-record exception can remove the single-source cap for an eligible stable fact.
Combines the adjusted atomic claims according to their importance.
A limit on one low-weight detail affects only that detail’s weighted contribution. It does not automatically impose the same maximum on the entire statement.
04
A worked example.
This hypothetical example shows the arithmetic. It is not a live fact-check.
PartWeightEvidence scoreContribution
Stable reference fact70%9566.5
Current-event detail, capped at 7430%90 → 7422.2
Final scoreF = round(66.5 + 22.2) = 89
05
Guardrails prevent false confidence.
Unverified sources may add context, but they cannot support a confident conclusion on their own.
One identity-verified authoritative source may be sufficient when it directly proves a stable fact and credible evidence does not contradict it.
A single announcement can establish that it was made, but not every factual claim within it.
Politics, safety, health, finance, crime, misconduct, and reputational allegations require stronger corroboration.
The system explains why the statement cannot be verified as a factual claim.
06
What the score cannot tell you.
The score is only as good as the evidence available at the time of the search. Sources can be incomplete, new information can emerge, and automated analysis can make mistakes. A high score is not a guarantee; a low score does not prove intent to deceive.
Common knowledge may have limited web evidence.
Some facts are so widely accepted, stable, or self-evident that few websites publish pages specifically documenting or verifying them. Because FakeNewsZero relies on publicly available, retrievable, and citable sources, these claims may sometimes receive a lower confidence score when direct supporting evidence is difficult to find. A lower score in this situation does not necessarily mean that the claim is false. It means that the available web evidence was not sufficient to justify a higher level of confidence.
Every result therefore shows its claim breakdown and sources. The number is the beginning of the explanation, not the end of it.
