Methodology

How we analyze each video, what we measure, and why.

Process

Each video is analyzed through a systematic, repeatable pipeline:

Transcript extraction — Auto-generated English subtitles downloaded via yt-dlp in SRT format with timestamps.
AI analysis — The full transcript is provided to Claude Opus 4.6 with the analysis schema and scoring rubric. The model reads the entire transcript and produces a structured JSON assessment, including claim verification with evidence.
Screencap extraction — For every timestamped item (quotes, rhetoric, claims), a video frame is captured automatically via ffmpeg.
Publication — JSON is rendered to static HTML. No editorial intervention beyond the schema design.

Scoring Rubric

Each video is evaluated on six axes, scored 1–5. Higher is better. Click any score on an analysis page to read the full justification.

Axis	What it measures	1 means	5 means
Factual Accuracy	Are stated facts, dates, figures, and causal claims correct?	Major errors	Solid throughout
Argumentative Rigor	Is reasoning logically sound? Are conclusions supported by the evidence presented?	Fallacious	Rigorous
Framing & Selectivity	Is the evidence cherry-picked? What's included versus excluded?	Highly selective	Balanced
Source Quality	Are claims grounded in credible, identifiable sources? Or vague appeals to authority?	Unsourced	Well-sourced
Perspective Diversity	Are competing interpretations acknowledged? Counterarguments engaged?	Single narrative	Multiple perspectives
Normative Loading	How much moral judgment is embedded? Is the audience told what to think?	Heavily prescriptive	Descriptive / analytical

Claim Verification

Every factual claim identified in a video is categorized and assessed:

Status	Meaning
verified	Claim is supported by credible evidence and mainstream expert consensus.
partially verified	Core direction is correct but details are wrong, exaggerated, or oversimplified.
disputed	Claim is contradicted by credible evidence or represents a fringe position.
unverifiable	Claim cannot be empirically tested, is speculative, or lacks sufficient evidence either way.

Limitations & Caveats

AI analysis is not infallible. LLMs can misidentify rhetoric, misjudge accuracy on niche topics, or miss cultural context. Every justification is published so readers can evaluate the reasoning.
Auto-generated transcripts contain errors. YouTube's speech-to-text occasionally garbles names, technical terms, and non-English words. Where a quote seems off, check the video.
Scores reflect the rubric, not a verdict on the speaker. A low score on one video does not characterize the entire channel.
This is content analysis, not mind-reading. We analyze what was said, how it was framed, and what was omitted. We do not speculate about the speaker's intentions.

Reproducibility

Every claim in every report is traceable to a timestamp in a public YouTube video. The analysis schema is published — anyone can feed the same transcripts and rubric to an LLM and compare results.

Fair Use Notice

This site constitutes fair use under Section 107 of the U.S. Copyright Act (17 U.S.C. § 107) and analogous provisions in other jurisdictions. All original video content remains the property of its creator(s).

The four statutory factors:

Purpose and character of the use. This project is transformative — it subjects publicly available videos to systematic critical analysis, commentary, and research.
Nature of the copyrighted work. The source videos are published on YouTube, a public platform.
Amount and substantiality. Only brief, timestamped quotations are reproduced to support specific analytical points. Individual video frames are used as contextual reference.
Effect on the market. This analysis does not compete with the original videos and is likely to drive additional viewership to the source channels.

All quoted material and screencaps are used solely for purposes of criticism, commentary, and scholarly analysis. Each item links or refers to the original public source.