How do engines understand video?
Engines understand video through the text around it: the title, the description, the chapter markers, the transcript, and the structured data on the page that embeds it. They can identify key moments from chapters and, in some cases, from speech, but the reliable signals are the ones you write.
For B2B video, which is usually explanatory, that is good news. An explanation has a natural structure of questions and answers, and that structure can be made explicit.
How should chapters be written?
Chapters should be titled as the questions they answer, in the same form as the H2s of a well-structured article: how does the approval gate work, what does a Digital FTE cost. Each chapter should open with the direct answer, so that a viewer landing on a key moment gets the response immediately.
This is why the SPEC-035 standard on ftes.ai requires that every piece's H2 set works as video chapters. An article and a video planned together share their structure, and the chapters are already written when the article is.
Why publish the transcript?
Publish the full transcript on the same page as the video because it is what engines retrieve and cite. A transcript, lightly edited for readability and broken under the chapter headings, turns a video into a set of self-contained text passages, which is exactly what AI search retrieves.
It also serves accessibility and readers who prefer text. A page with an embedded video, question-headed chapters, and the transcript beneath is one page that serves viewers, readers, and engines.
- Full transcript, edited for readability, on the video's page
- Broken under the chapter headings as H2s or H3s
- Each chapter's text opens with the direct answer
- Key terms and entity names written consistently with the rest of the site
How should VideoObject schema be used?
VideoObject schema should declare the real video embedded on the page: its name, description, thumbnail URL, upload date, duration, and content or embed URL, with optional clip markers for the chapters. It should be server-rendered so that crawlers that execute no JavaScript see it.
It should never be added to a page with no video. Schema that describes content not visibly on the page violates engine guidelines and risks the page's ranking. The same rule applies on ftes.ai to FAQPage schema: only where the FAQ is visible, only for a video that is really there.
How do you plan video and article together?
Plan them together by writing the article first to the standard, with its fan-out list of sub-questions and its question H2s, then recording the video to that structure, chapter by chapter. The transcript will then match the article closely, and the two reinforce each other on the page rather than competing.
Measure video visibility by sampling the target questions and recording whether the video appears as a result or key moment, alongside the usual page-level measurement. A Search Visibility FTE can plan the article and its chapter structure; the recording is a human task.
What does the planned video-plus-article page look like?
As a labelled example, a page about how a Finance Controller FTE works carries the article written to the standard at the top: a 40 to 60 word direct answer, question-headed sections on what it owns, how matching works, what happens with exceptions, what earns autonomy, and how the audit pack is built, and a short FAQ block. Below or beside it sits the embedded video recorded to the same structure, with chapters titled by those same questions and each chapter opening with the direct answer, and beneath the video the full transcript broken under the same headings.
The page's JSON-LD carries BlogPosting and FAQPage for the article and VideoObject for the real video, with the chapters as clip markers, all server-rendered. A viewer landing on a key moment gets the answer immediately; a reader gets the same answer in text; an engine gets three consistent representations of one set of passages. The recording is a human task, but the plan, the chapters, and the article are what a Search Visibility FTE produces, and the measurement afterwards samples the questions for both the page and the video.