Our Methodology

ClarityCheck is a reasoning aid, not a truth oracle. We evaluate the strength of publicly available evidence — so you can make more informed judgments about the claims you encounter.

How the Evidence Score Works

When you submit a claim, our AI assistant searches multiple sources, classifies the claim type, evaluates the credibility and stance of each source, and synthesizes an evidence strength score from 0 to 100.

81–100

Well-Supported (Strong)

Overwhelming evidence from multiple credible sources supports this claim.

51–80

Well-Supported (Moderate)

Majority of credible evidence supports this claim, with minor caveats.

21–50

Contested

Evidence is mixed. Credible sources exist on both sides of the claim.

1–20

Weakly Supported

Very little credible evidence supports this claim. Most evidence opposes it.

Unverifiable

Insufficient publicly available evidence to evaluate this claim either way.

Source Weighting

Not all sources are weighted equally. Our system considers source type, credibility tier, and the consistency of evidence across sources.

Primary Research

Highest

Peer-reviewed studies, original datasets, first-hand documentation

Government/Institutional

High

Official reports from government agencies, WHO, CDC, courts

Academic

High

University research, systematic reviews, meta-analyses

Quality News

Medium

Major wire services, established investigative journalism

Aggregator/Secondary

Lower

Fact-check sites, encyclopedias, secondary reporting

Ethical Guardrails

We take responsibility for how evidence analysis can influence thinking. These are the non-negotiable rules built into every analysis.

No Truth Claims

Scores measure evidence strength, not absolute truth. We never claim to know "the truth."

Both Sides Always

Every analysis shows the strongest argument for AND against, regardless of where evidence leans.

No Fabricated Sources

Only sources actually retrieved are cited. No synthetic or hallucinated citations.

Neutral Framing

Political and social topics use neutral language. No partisan source selection.

Opinion Detection

Pure opinions, value judgments, and future predictions are flagged, not scored.

Paraphrased Only

Source summaries are paraphrased. Maximum one short quote (<15 words) per source.

What ClarityCheck Is NOT

• Not a truth oracle. We measure evidence strength, not absolute truth. The score may change as new evidence emerges.

• Not a replacement for expert judgment. Complex scientific, medical, or legal claims require domain expertise beyond what any automated tool can provide.

• Not immune to AI limitations. While we have guardrails against fabrication, no AI system is perfect. Always verify critical claims with primary sources.

• Not a political tool. We do not take sides. Neutral framing and balanced source selection are core principles.