How We Score
An independent, proprietary scoring system for YouTube video quality. Evidence-gated.
Proprietary by Design
Our scoring system is proprietary. We don't publish weights, models, or specific thresholds. What we do publish: what gets measured, what counts as evidence, and what has to pass before a score is released.
Things You Can Verify
We only score signals you could check yourself if you had the time:
- •What someone actually said in the video
- •Sources they cited or referenced
- •Whether their statements are consistent
- •Sponsorship disclosures
- •Factual accuracy where we can verify it
When we measure sentiment, it is always about a specific statement regarding a specific brand or product.
What's In, What's Out
We analyze what's actually in videos. Spoken words, context, who and what gets mentioned.
We record basic metadata like publish date and view count for context, but engagement numbers don't factor into our scoring. Popularity isn't quality.
Apples to Apples
Videos are scored against others of the same kind. Podcasts against podcasts. Product reviews against product reviews. News briefs against news briefs. Analysis against analysis.
A score in one content type isn't directly comparable to a score in another. Each has its own standards.
Everything Traces Back
Every finding traces back to something someone actually said. No exceptions.
You can see the exact quotes. If we can't point to what was said, we don't publish it.
Example
What Has to Pass
Not every video makes it through. Here's what we check:
Clear Audio
We have to be able to hear what's being said. If the audio is garbled or incomplete, we can't analyze it.
Enough Evidence
We need a minimum number of quotable statements before we'll publish a finding. One line isn't enough.
Confidence Check
If we're not confident, we flag it. Sometimes we hold it back entirely.
Safety Review
Content flagged for safety concerns doesn't get published.
Published vs Withheld
We publish when
- Video meets all standards
- Findings have enough evidence
- Scores have clear justification
- Everything traces to what was said
We hold back when
- •We can't hear what's being said
- •There's not enough evidence
- •Content is flagged for safety
- •We're not confident and can't explain why
See It in Action
Try the extension on any YouTube video, or explore our enterprise data feed.