What Does Video Structured Data Do for SEO and AI Visibility?

What does VideoObject markup do in Google?

Google's video documentation sets out the basics:

• Required properties: name, thumbnailUrl and uploadDate.
• Strongly recommended: contentUrl, the video file itself, which Google calls the most effective way for it to fetch your video. If that is not available, use embedUrl for the player. Also add description and duration.
• Optional enhancements: Clip or SeekToAction for key moments, BroadcastEvent for live streams, and interactionStatistic for view counts.
• A watch page. The markup must be on a page where users can watch the video. For video features, Google looks for a dedicated watch page whose main purpose is showing that video, and the page itself must be indexed.
• Stable thumbnails. Thumbnail URLs that change often can stop videos being indexed.

Markup does not guarantee video features. It makes the video eligible and easier for Google to process.

Do WebVTT captions and VideoObject schema improve AI video indexing?

They help in different ways, and neither does the whole job.

VideoObject schema helps Google find and index the video, which matters for Google's AI features. They draw on Google's index, and video results and YouTube are cited often in AI Overviews. For other engines, the evidence is weaker. Tests of ChatGPT, Claude, Perplexity and Gemini fetching pages directly found they ignored JSON-LD and read visible text. So the schema's name and description help those systems only if the same information is also visible on the page.

WebVTT captions make the video accessible and let players show subtitles. Google's video documentation does not describe using caption tracks from web pages for indexing, and AI companies do not document fetching them. YouTube is different: its captions are part of the platform Google indexes, and Gemini can work with public YouTube videos directly.

The dependable route is to publish the same words as HTML:

• A full transcript, or a detailed summary with the key points, figures and names.
• A chapter list with timestamps and descriptive headings.
• A description that says what the video covers and who it is for.

Do LLMs read WebVTT and SRT transcripts to understand videos and podcasts?

It depends on where the file is:

• On your website, as a <track> file for a video player: no AI company documents fetching or using these files. Treat them as invisible to AI crawlers and publish the transcript text on the page.
• On YouTube: captions are part of the video's data on the platform. Upload accurate captions rather than relying on automatic ones, especially for product names and technical terms.
• In podcast feeds: Apple Podcasts generates transcripts automatically and also accepts creators' own transcripts through the RSS transcript tag, in VTT or SRT format. That improves the transcript listeners see in Apple's app. It does not put the text on the open web for AI crawlers.
• When a user uploads the file to a chat assistant, the model can read it like any other text.

For podcasts, the same rule applies as for video. Publish an episode page with show notes and a transcript in HTML, naming guests, companies and key claims.

Does SeekToAction markup help AI systems cite specific video sections?

It helps Google show key moments. There is no documented effect on AI answers.

SeekToAction tells Google how your video URLs express a start time, for example ?t= followed by seconds. Google can then identify key moments automatically and link users straight to them. It works only if Google can fetch your video files. The alternative, Clip markup, lets you define each segment's start and end times and labels yourself. Google says it prioritises key moments set by you, through structured data or YouTube chapters. Its own automatic detection works without markup.

AI answers that cite video usually link to the video, and occasionally to a timestamp. To make a specific section citable:

• Give each important section its own heading and timestamp in the page's chapter list.
• Put the section's key point in text under that heading.
• Use timestamp links such as ?t=245 in the chapter list, so readers and systems can reach the moment.
• Consider separate short clips on their own watch pages for sections people ask about often.

What should I include in a video sitemap for AI search discovery?

A video sitemap helps Google find videos, especially those that are hard to discover through links. Google's documentation lists these tags:

• Required: video:thumbnail_loc, video:title and video:description, the last up to 2,048 characters and matching the page's description. Also required is either video:content_loc, the media file and Google's preferred option, or video:player_loc, the player URL, which is used for embeds from platforms such as YouTube and Vimeo.
• Useful optional tags: video:duration (1 to 28,800 seconds), video:publication_date, video:expiration_date for videos that will be removed, video:family_friendly, video:restriction for country limits, and video:requires_subscription.

The content URL must not be the same as the page URL. The title and description should match what the page shows. Google also accepts media RSS feeds as an alternative.

Google and Bing both read video sitemaps, and Bing also accepts mRSS feeds, so submit the sitemap in Search Console and Bing Webmaster Tools; Bing's index feeds Copilot. Other AI engines discover pages through their own crawlers, which follow links and standard sitemaps. So the watch page itself must be linked from your site and listed in your main sitemap.

Where should the transcript live, and how should it look?

• On the watch page, below the player, as HTML text, not in a collapsed element that loads by script.
• Edited for reading: remove filler words, fix names and add headings at chapter breaks.
• With speaker names for interviews and panels.
• With key figures and claims stated exactly, since those are what answers quote.

If a full transcript is too long, publish a detailed summary with the key points under chapter headings, and link the full transcript. Whether embedding videos in articles helps is a related question, covered in whether embedded videos in blog posts improve SEO and AI visibility.

What does a complete video page include?

For YouTube videos, apply the same thinking on the platform, as covered in how to optimise YouTube videos so AI answers cite them.

What might an illustrative case look like?

Consider an illustrative software company with 60 product tutorial videos embedded on help pages, marked up with VideoObject but with no text beyond a title. Google showed some of the videos in results, but AI answers to "how do I set up [feature] in [product]" cited a competitor's written guides. The company added edited transcripts and chapter lists to each tutorial page, and Clip markup for the longest videos. It also listed the watch pages in its sitemap. Over the following two months, ChatGPT and Perplexity began citing its tutorial pages for setup questions, quoting steps from the transcripts.

What do people ask about video structured data and AI?

Is VideoObject markup still worth adding?

Yes, for Google's video features and indexing. Just do not expect it alone to change what other AI assistants say.

Should the transcript be in the markup too?

Schema.org has a transcript property, but Google does not list it. Put the transcript in the visible page.

Do automatic captions count?

They help accessibility, but errors in names and terms carry into anything built from them. Correct them for important videos.

Does hosting on YouTube or our own site matter more?

Both have value. YouTube is heavily cited in Google's AI answers; your own watch pages let every engine read your transcript.