How we make your report
Our AI watches each YouTube review in full to identify the product and what the reviewer says about it. AI writes the report’s key conclusions from that evidence, and a person listens to every quoted line before you see it.
Three passes through each review.
Map the video
Identify products, speakers, sections and disclosed commercial relationships. Distinguish exact models and variants, brand-only mentions and products outside the catalogue. Speakers are named only when the video identifies them.
Examine sections
Examine video sections of approximately 2.5–6 minutes with captions. Keep claims with the surrounding reasoning, conditions and timing, such as “out of the box” or “after a week”.
Check and connect
Build reviewer verdicts with their reasons and caveats, revisit checkable claims, and match quotations to caption locations. Checks can support or correct a claim, or mark it unsupported by the words.
The distinctions behind a finding.
The analysis keeps the distinctions that matter when you interpret a reviewer’s conclusion. Not every review contains every kind of evidence.
The report brings the relevant findings together to answer your questions. Quotations and timestamped links let you examine the words behind them. Implications and possible next steps are distinguished from reviewer statements.
- Recommendations and comparisons
- Preserve what a reviewer recommends, for whom and why. Distinguish explicit preferences, ties, conditional or hypothetical judgments, and differences without a preference.
- Experience and context
- Show whether a claim comes from reported use, a measurement, specifications or secondhand information, with the product, speaker and conditions attached.
- Reasons, caveats and open questions
- Keep supporting reasons, contradictions and changes of opinion connected to the conclusion. Retain what the reviewer explicitly leaves untested.
- Commercial relationships
- Record sponsorship, gifted products, affiliate links, discount codes and ad reads separately. They have different meanings and may concern different products.
Keep the limits in view.
The evidence helps frame decisions and questions for investigation. These limits remain part of its interpretation.
- Reviewers and coverage
- The analysis describes reviewer statements, not buyer sentiment or purchasing behaviour. Findings depend on available reviews; missing evidence remains unknown.
- Words and automated checks
- Caption matches locate words, not objective product truth. Automated checks are not comprehensive human verification, and we do not independently test product performance.
- Disclosures
- An absence of recorded disclosure does not establish independence or rule out a commercial relationship.
- Video views
- Views do not establish unique reach, sales or conversion.
Earlier public examples
These examples use our earlier topic-counting method. Interpret their figures using the historical method shown on each page. The current approach above examines recommendations, reasons and caveats.
Separate market research
Our search and YouTube market dataset is a separate research product. It does not feed the product-review reports.
How the market data worksBring your product and your question.
Share your product, relevant rivals and a question. We assess review coverage, then reply with what it can support and a scope to discuss.
or email contact@sourcesignal.io