Video SEO

Search engines cannot watch a video the way a person does. What they index is the text around it: the title, the description, chapters, captions and transcript, the structured data, and the page the video sits on. Video SEO is getting that text right, so the video can be found in search results, in video carousels and, increasingly, in AI answers.

What does Google need before a video can appear?

Google's video guidance comes down to a short list:

A page whose main purpose is something else, with a video tucked halfway down, is weaker for video results than a dedicated watch page.

Do AI answers cite videos?

They do, less than people expect, and the amount depends on the engine and the category. Attensira's documentation lists threads, YouTube and roundups as what tends to decide answers on Perplexity, Gemini and Google's AI Overviews, and a brand's own docs and pricing pages on ChatGPT. That is the default before a workspace has its own evidence.

Here is our own evidence, for our category. Across 4,524 ChatGPT and Perplexity answers to the 61 questions Attensira tracks about AI search, in the 30 days to 4 October 2026, read over the Attensira MCP:

• youtube.com was not among the 25 most-cited domains. The 25th had 422 of the 100,380 citations, so YouTube had fewer: under half of one percent.
• It was a deciding source for 1 of the 61 questions, counting a domain as deciding when it is among the six ranked as deciding a question's answers on either engine.

That one question is where the picture flips. Asked "why am I not showing up in ChatGPT", Perplexity's 54 answers over the same 30 days gave youtube.com 96 of 535 citations, more than any other domain ranked as deciding them, OpenAI's help centre included. One question can be decided by video even when its category is not.

Across a category of software tools as a whole, text wins. In categories where people learn by watching, such as cooking, repairs or fitness, the share can be very different, which is why it should be measured per category and per engine rather than assumed.

What makes a video citable in an AI answer?

The same things that make text citable, applied to the text the video carries:

• A title that states the question the video answers.
• A description that answers it in the first lines, not a list of links and hashtags.
• Chapters with descriptive names, so a passage can be matched to a part of the video.
• A transcript or accurate captions, published where engines can read them.
• A watch page on your own site with the transcript and VideoObject markup, so the video's content exists as indexable text you control.

What do people get wrong about video SEO?

• Optimising the video file. Engines index the metadata and the page.
• Publishing on YouTube only. You get YouTube's audience, but not a watch page you control.
• Writing descriptions for the algorithm. Keyword lists and hashtags tell an AI engine nothing it can quote.
• Duplicating metadata across videos. Every video needs its own title and description.
• Blocking thumbnails or files in robots.txt. A video without a fetchable thumbnail is not eligible.

How does Attensira help with video?

Attensira records every link each answer cites, so a YouTube video that decides one of your tracked questions shows up by name in the deciding-sources map for that question and engine, marked as yours, a competitor's or someone else's. Until your workspace has at least 20 successful reads on a source, the next move comes from a static per-engine table, which names YouTube among what decides answers on Perplexity, Gemini and Google AI Overviews; after that it comes from your own history. When someone else's video decides the answer, the move is to be in it or to publish the video that answers the question better; the metadata checklist above is what makes yours readable. Inside a project, that work can go to the crew's Video role. See YouTube visibility for how YouTube is treated as a place answers are built from.