HowTrue listens to a video and scores it — not just whether the claims hold up, but whether the speaker is arguing honestly or working you.
A claim can be technically accurate and still be used to mislead you. HowTrue scores both axes and won't call a video trustworthy if either one fails.
Every checkable claim is pulled from the transcript and tested against the evidence — sourcing, currency, and whether the claim survives its own strongest counterexample.
The same transcript is read for the moves a persuader makes when the facts are thin — the fallacies, the framing tricks, the appeals designed to end your thinking rather than inform it.
HowTrue pulls the full spoken transcript from the video, timestamped line by line.
Factual claims are lifted out; rhetorical moves are flagged where they occur.
Each claim is weighed on sourcing and accuracy; each tactic on how much it distorts.
Both axes collapse into one HowTrue score — and roll into the speaker's running record.
One dishonest video is a data point. HowTrue keeps every audited speaker's track record, so a channel that earns your trust — or loses it — carries that history into the next thing they post.
Consistent sourcing across 22 audited talks. Occasional overstatement, always corrected on record. Score reflects the pattern, not a single good day.
High-confidence claims, low-quality sourcing, heavy reliance on urgency and outrage framing. Nine audits, same shape every time.
Pick the veracity range you'll allow into your feed. Set a minimum to keep the noise out — or invert it to isolate exactly the low-scoring content you want to examine.
The filter takes a minimum and a maximum. Most people set a floor around 60 and stop seeing the junk. Researchers, journalists, and the merely curious flip it the other way — pin the max low and surface only the manipulation, all in one place.
Drag the handles. The list below re-sorts to whatever range you leave them at.
HowTrue rates the veracity of what was said in a specific video, using a method we publish. It measures claims and tactics — it does not pronounce on anyone's character, and every score links to the transcript lines and reasoning behind it, so you can check our work.
Speakers can respond, and audits are re-run when a video is updated or a claim's evidence changes. We'd rather be transparent and contestable than confident and closed.
A triage read is always free. A full audit — every claim tested, every tactic flagged, the speaker's record updated — costs one credit.