About Held True
Mission
Long-form podcasts shape how millions of people understand the world. A single three-hour episode can contain hundreds of factual claims that go unchecked. Held True exists to change that.
We analyze YouTube podcasts claim by claim: every verifiable statement is extracted, checked against primary sources, and timestamped so you can see exactly when it was said.
How It Works
Every verifiable statement in a video is identified and timestamped. Each claim is then checked against primary sources: peer-reviewed studies, official data, government reports, and authoritative references. Based on the evidence found, each claim receives one of seven verdicts:
| Verdict | Meaning |
|---|---|
| true | Factually correct and verified by evidence |
| inexact | Core assertion holds but contains imprecisions in secondary details |
| false | Core factual assertion is contradicted by evidence |
| outdated | Was true at some point but no longer accurate |
| unsubstantiated | Claim made without any evidence or source |
| disputed | Credible sources contradict each other, no clear consensus |
| unverifiable | No evidence found to confirm or deny |
Use of AI
Held True uses advanced language models to assist with claim extraction and verification. The AI is a tool, not the authority. The sources are.
Every verdict is accompanied by its source references so you can verify the conclusion yourself. We believe transparency about methodology is more valuable than pretending to be infallible.
AI can make mistakes. That's precisely why every claim links to its sources.
Corrections
If you believe a verdict is incorrect, we want to hear about it. Accuracy is the entire point.
Contact us at with the video, the claim, and your evidence. We review every correction and update verdicts when warranted.
Our policy: correct, never delete.
Contact
For corrections, feedback, press inquiries, or partnership proposals: