Most brand films are delivered, posted once, embedded on a homepage, and left there. The search value is thrown away, and it is significant.
The reason is a misunderstanding about what a machine sees when it hits a video.
Machines Do Not Watch, They Read
A crawler encountering a video element sees a file reference. It does not know what is in the frame, who is speaking, or what the brand does. A generative engine like ChatGPT or Perplexity is in the same position.
What both can read is everything around the video: the page copy, the transcript, the caption track, the schema, the filename, and the platform description. That surrounding text is the entire search surface of a video asset.
This is good news, because that text is also the densest, most specific language a brand ever produces. Somebody on camera explaining exactly what the business does, in plain speech, is precisely the kind of extractable answer generative engines favor.
The Five Things That Turn a Film Into a Search Asset
- A real transcript on the page. Not a hidden div stuffed under the fold, an actual readable transcript or a well-written summary built from it. This is the single highest-value step and the one almost nobody does.
- A caption file. SRT or VTT, uploaded to the platform and attached to the player. Captions serve accessibility first and indexing second, and both matter.
- VideoObject schema. Name, description, thumbnailUrl, uploadDate, duration, and contentUrl or embedUrl. This is what makes the page eligible for video rich results and what labels the asset for machine readers.
- A deliberate thumbnail. A pulled frame from the graded master, not an auto-selected frame of somebody mid-blink. It is the image search results and social previews will use.
- A descriptive filename and page. The video should live on a page that is about the same subject the video is about. A film about hospitality production embedded on a generic homepage wastes the topical signal.
The Publishing Split
Publish to both surfaces, for different jobs.
YouTube is a discovery engine with its own index, and its transcripts get crawled and reused widely, including by generative engines. It also functions as a second entity signal: a consistent brand presence on another high-trust surface, which is exactly the kind of corroboration AI engines weigh when deciding who to recommend.
The site version carries the conversion job and the schema. It should be the shorter cut, sized and compressed for web delivery, with the transcript and structured data attached.
Why This Has To Be Scoped Before the Shoot
Every one of those five items is cheaper to produce at capture time than to retrofit. The transcript comes from the production audio, which means the audio has to be captured cleanly. The thumbnail comes from the graded master, which means somebody has to pull it before delivery. The aspect ratios each destination needs have to be framed for on set.
A studio that shoots and then hands the file to a separate web vendor cannot do any of that, because the decisions were made before anyone asked the question. This is the practical case for one roof.
UM Media scopes production deliverables and search structure in the same engagement. See the SEO / AEO / GEO Foundation, or Visual Production. Structure only, no ranking or citation guarantees.