How does TubeLens evaluate a YouTube video?+
We read the full transcript of the video and score it 0–10 across four dimensions — information density (weight 30%), clarity (30%), credibility (30%), and originality (10%). In parallel, we detect up to 28 quality signals in the content, each with an intensity of 1–5 and a justification citing a transcript excerpt as evidence.
How long does an analysis take?+
Typically 5–15 seconds per video after the URL is submitted. Videos analyzed before (globally cached by video_id) respond instantly, without reprocessing or burning fresh tokens.
Why does the channel ranking use a Bayesian average?+
Because a channel with 2 videos at score 10 should not statistically beat a channel with 20 videos at score 9.2. Bayesian smoothing with the global mean as prior fixes this via (C × M + n × x) / (C + n) with C = 5. Channels with fewer than 3 analyzed videos stay out of the ranking.
What signals does TubeLens detect in videos?+
28 patterns: 15 red flags (pseudo-scientific, conspiracy theorist, sensationalist, clickbait, misinformation, covert advertising, charlatanism, dogmatic, rage bait, etc.), 6 neutral (satire, disclosed opinion, speculative, controversial topic), and 8 green flags (well-sourced, didactic, original, balanced, transparent, rigorous, in-depth, up-to-date).
Does TubeLens work on videos without captions?+
Not yet. The analysis depends exclusively on the text transcript, so the channel needs to have CC enabled — manual or YouTube auto-captions. If a video has no captions available, we show a clear message rather than fabricating an analysis.
Is TubeLens free?+
Yes. Sign-up is free and each user can analyze as many videos as they want. Videos already evaluated by others show up instantly — every analysis becomes a public cache entry keyed by video_id.
