Why separate summaries are not a comparison
Two summaries can accurately describe two videos and still hide the central disagreement. A useful comparison needs shared dimensions: the same question, the same definition, and the same standard of evidence.
YouTube Conversation can retrieve the complete available timestamped transcript for each public captioned video. ChatGPT can then compare the spoken arguments while preserving links back to each source.
The comparison workflow
- Choose two to five videos that address the same real question.
- Retrieve each transcript and create a short source ledger with title, speaker, date, and URL.
- Define three to seven comparison dimensions before judging the answers.
- Fill the matrix with claims, evidence, assumptions, and timestamps from each video.
- Write a synthesis that distinguishes consensus, disagreement, and unresolved uncertainty.
VIDEO_URL_1
VIDEO_URL_2Compare these videos on the same question. Build a matrix covering each speaker's core position, definitions, strongest evidence, assumptions, caveats, and unanswered questions. Link every substantive point to the exact timestamp in its source. Finish with the most important genuine disagreement, not a generic summary.
Comparison dimensions that reveal something
- Problem definition: Are the speakers answering the same question?
- Causal model: What does each person believe causes the outcome?
- Evidence: What data, examples, or sources support the argument?
- Assumptions: What must be true for the conclusion to hold?
- Tradeoffs: What cost or risk does each position accept?
- Falsifiability: What future evidence could change the speaker's mind?
A synthesis is only as useful as your ability to inspect it. Keep every important conclusion attached to the exact source moment that produced it.
Prompts for deeper comparison
- “Where are they using the same word to mean different things?”
- “Which disagreement is factual, and which is a difference in values?”
- “What evidence would both speakers accept?”
- “Whose conclusion depends on the most uncertain assumption?”
- “Create the strongest version of each position before evaluating either.”
Limits to keep visible
V1 works from available timestamped transcripts. It can compare what was spoken, but it does not automatically interpret silent charts, gestures, or demonstrations that were never described aloud. It also does not make a claim true merely because multiple speakers repeat it.
For a claim-by-claim audit before comparison, use the guide to fact-checking a YouTube video with ChatGPT.