Skip to content

Reconcile what a video says with what it shows

Review spoken and visible evidence separately when they appear to disagree, and preserve the limitations of any available video analysis. Record unresolved differences.

Guides

How should a reviewer handle conflicting narration and imagery?

A demonstration may describe a phone stand as stable while the visible device shifts during adjustment. Neither narration nor imagery should automatically erase the other. A useful review identifies the specific spoken claim, the observable scene and the conditions under which they appear, while avoiding conclusions about hidden tests or physical properties not actually shown.

Example research prompt
Review the supplied phone stand demonstration for differences between narrated stability claims and visible adjustment behavior. Separate observed moments from publisher-note-only answers.
01 / Guides

Specify the claim and moment

Start with one public video link and a precise question about the adjustment sequence. Include visible movement relevant to the stated stability claim; exclude unrelated camera shake or edits unless their effect can be assessed. A scene description should state what is visible rather than diagnosing why it happened.

02 / Guides

Inspect the available analysis

Use watch_video for timestamped spoken and visible evidence when supported. If analysis is pending, follow the returned retry guidance. If the response is based on publisher notes, do not describe it as visual inspection. Preserve the answered_from limitation so a polished summary does not quietly upgrade descriptive text into observed footage.

03 / Guides

Compare without overclaiming

Place the spoken statement and relevant moment side by side. Consider whether they concern different conditions, such as resting versus adjustment. If an edit prevents continuity or the frame is unclear, keep the discrepancy unresolved. Do not infer a measured stability rating from a short demonstration that provides no defined testing protocol.

04 / Guides

Deliver an evidence distinction

Summarize what was said, what was visible and what cannot be determined. Include timestamps only when returned evidence supports them. If video analysis is unavailable, hand off a restricted review of available publisher material or request another permitted source, rather than inventing observations to finish a multimodal comparison.

Keep the boundaries clear.

  • Publisher notes cannot substitute for a claim that frames were inspected.
  • Examples describe review methods, not live tests or guaranteed platform coverage.

Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.

Primary references

Documentation reviewed 2026-11-13. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.

Know what social media shows about you.
Start with a question.

Get started
montage