Reacher's AI analysis of a single video: the spoken transcript, a hook_summary of how it opens, the structure_outline of its beats, the on-screen text (OCR), the visual tags, the products detected on screen, and the scored creative features (hook in the first three seconds, product demo present, voiceover present, pace, and a content summary).

video_id is the value returned by POST /videos/list, POST /videos/performance and GET /crm/content.

Coverage is partial and grows over time: a video this shop owns that has not been analyzed yet returns 200 with analyzed: false and null fields, so absence is distinguishable from a bad id. Thumbnail-source analyses (analysis_source: thumbnail) carry no transcript and no motion features.

The analysis returned is of the creator's publicly-posted video content — public content anyone can watch, not another customer's data. Any analyzed video is readable here: one of your own shop's posts returns the full response; any other analyzed video returns the same public creative fields with no shop-derived metrics. The is_own_content flag says which (true = your own post, false = the creator's public content that is not one of your posts). A false value never means the video belongs to another customer, and no other shop's identity, metrics, or ownership is ever disclosed. Only a video_id with no analysis at all returns 404.

Set include_experimental=true to additionally receive features_experimental, the analyzer's full scored-feature block — richer, but unstable across analyzer versions.

Single-shop only — set x-shop-id to one shop ID rather than all or a comma-separated list.