1. Claim Extraction

We use AI to transcribe podcast episodes and identify verifiable factual claims — statements about data, events, science, or history that can be checked against evidence. Opinions, predictions, and subjective statements are excluded.

2. Evidence Gathering

Each claim is cross-referenced against multiple authoritative sources:

• Peer-reviewed scientific papers and meta-analyses
• Government data (BLS, CBO, Fed, CDC, WHO)
• Established news organizations with fact-checking teams
• Academic institutions and expert consensus
• Primary sources cited by the speaker

3. Verdict Scale

• True — Fully supported by available evidence
• Mostly True — Largely accurate with minor caveats
• Mixed — Contains both accurate and inaccurate elements
• Mostly False — Significant inaccuracies despite some truthful elements
• False — Contradicted by available evidence

4. Accuracy Grades

Each speaker receives a letter grade based on weighted accuracy: True=100%, Mostly True=80%, Mixed=50%, Mostly False=20%, False=0%. Grades: A ≥ 85%, A- ≥ 80%, B+ ≥ 75%, B ≥ 70%, B- ≥ 65%, C+ ≥ 60%, C ≥ 55%, D ≥ 40%, F < 40%.

5. Transparency

Every verdict includes the original quote, context, sources consulted, and reasoning. We show our work so you can evaluate our conclusions independently. PodCheck is not the arbiter of truth — we surface evidence and let you decide.