VideoObject Schema: Making Videos AI-Discoverable and Citable

What Is VideoObject Schema?

VideoObject Schema is a schema.org structured data type that provides machine-readable metadata about a video, including its title, description, thumbnail image, duration, upload date, content URL, and publisher. This structured data enables search engines to display video rich results (video thumbnails in search listings) and helps AI models understand what a video covers so they can cite it in responses.

Video content is inherently opaque to text-based AI models -- they cannot watch a video to determine its content. Instead, they rely on metadata: the title, description, transcript, and structured data. VideoObject Schema provides the most reliable, structured source of this metadata. Without it, AI models must infer video content from surrounding page text, which is less accurate and less likely to result in a citation.

For general structured data background, see our JSON-LD basics guide. For the broader strategy of video optimization for AI, see our guide on YouTube optimization for AI search.

Why Video Matters for AI Search

Video is an increasingly important content format for AI citations. Several data points illustrate why VideoObject Schema should be part of your AI SEO strategy.

Video citation statistics

• Perplexity cites YouTube videos in 16.1% of responses -- making YouTube the most-cited non-traditional web source
• Google's AI Mode and AI Overviews increasingly include video thumbnails in results
• ChatGPT's browsing capability can access video metadata from pages with VideoObject Schema
• Video content is growing in AI citation share as multimodal AI models improve

Why AI cites video

AI models cite videos for three primary reasons:

1. Tutorial and how-to content -- Videos demonstrating processes are cited when users ask "How do I..." questions
2. Product demonstrations -- Video reviews and demos are cited in product recommendation queries
3. Expert authority -- Videos featuring recognized experts carry E-E-A-T signals that strengthen citations

The discovery challenge

Unlike text articles that AI can fully parse, video content requires metadata to be discoverable. VideoObject Schema is the mechanism that bridges this gap -- it tells AI what the video is about, how long it is, when it was published, and where to access it.

Complete JSON-LD Implementation

Here is a complete VideoObject Schema implementation:

{
  "@context": "https://schema.org",
  "@type": "VideoObject",
  "name": "How to Add Schema Markup to Your Website in 10 Minutes",
  "description": "Step-by-step tutorial showing how to add JSON-LD Schema markup to any website. Covers Organization, Article, and FAQPage Schema with copy-paste code examples.",
  "thumbnailUrl": "https://aimetrico.com/images/video/schema-tutorial-thumb.webp",
  "uploadDate": "2026-03-15",
  "duration": "PT10M32S",
  "contentUrl": "https://aimetrico.com/video/schema-tutorial.mp4",
  "embedUrl": "https://www.youtube.com/embed/abc123xyz",
  "publisher": {
    "@type": "Organization",
    "name": "AImetrico",
    "logo": {
      "@type": "ImageObject",
      "url": "https://aimetrico.com/logo.png"
    }
  },
  "interactionStatistic": {
    "@type": "InteractionCounter",
    "interactionType": "https://schema.org/WatchAction",
    "userInteractionCount": "15400"
  }
}

Duration format

Duration uses ISO 8601 format: PT (period time) followed by hours (H), minutes (M), and seconds (S). Examples:

• 10 minutes, 32 seconds: PT10M32S
• 1 hour, 5 minutes: PT1H5M
• 45 seconds: PT45S

Required and Recommended Properties

| Property | Required | AI Value | Description | |---|---|---|---| | name | Yes | Critical | Video title -- should match the visible title | | description | Yes | Critical | 100-300 character description of video content | | thumbnailUrl | Yes | High | Preview image URL (minimum 60x30px) | | uploadDate | Recommended | High | Publication date (ISO 8601) | | duration | Recommended | High | Video length (ISO 8601 duration) | | contentUrl | Recommended | Medium | Direct URL to the video file | | embedUrl | Recommended | Medium | URL for embedding (YouTube embed URL) | | publisher | Recommended | High | Organization that published the video | | interactionStatistic | Optional | Medium | View count and engagement data | | transcript | Optional | High | Full text transcript of the video | | hasPart | Optional | High | Clip objects defining video segments | | inLanguage | Optional | Medium | Language code of the video content |

The transcript property

The transcript property is especially valuable for AI SEO. It provides AI models with the full text content of the video, making the video's actual content searchable and citable. Without a transcript, AI relies solely on the title and description.

Clip Markup: Timestamps for Key Moments

Clip markup allows you to define named segments within a video, each with a start time and end time. This enables Google's key moments feature in search results and helps AI models reference specific portions of a video rather than the entire piece.

Implementation

{
  "@type": "VideoObject",
  "name": "Complete AI SEO Guide",
  "hasPart": [
    {
      "@type": "Clip",
      "name": "What is AI SEO?",
      "startOffset": 30,
      "endOffset": 120,
      "url": "https://aimetrico.com/video/ai-seo-guide?t=30"
    },
    {
      "@type": "Clip",
      "name": "How to check robots.txt for AI crawlers",
      "startOffset": 120,
      "endOffset": 300,
      "url": "https://aimetrico.com/video/ai-seo-guide?t=120"
    },
    {
      "@type": "Clip",
      "name": "Adding Schema markup step by step",
      "startOffset": 300,
      "endOffset": 540,
      "url": "https://aimetrico.com/video/ai-seo-guide?t=300"
    }
  ]
}

Why clips matter for AI

When a user asks an AI model "How do I check robots.txt for AI crawlers?", the AI can cite the specific clip that answers this question rather than the entire 15-minute video. This increases the relevance of the citation and makes it more useful to the user.

VideoObject for YouTube Embeds vs. Self-Hosted Video

The implementation differs slightly depending on whether your video is hosted on YouTube or self-hosted on your own server.

YouTube embedded videos

When embedding a YouTube video on your page, add VideoObject Schema with:

• embedUrl set to the YouTube embed URL (e.g., https://www.youtube.com/embed/abc123)
• thumbnailUrl pointing to either your custom thumbnail or the YouTube-generated one
• Your domain's publisher information (not YouTube's)

This associates the video with your domain in structured data, helping AI models attribute the video to your brand when citing it from your page.

Self-hosted videos

For videos hosted on your own server:

• contentUrl set to the direct video file URL
• thumbnailUrl pointing to your generated thumbnail
• Include encodingFormat (e.g., "video/mp4") if relevant

Both platforms

Regardless of hosting, always include:

• A descriptive name that matches your page's visible video title
• A detailed description summarizing the video's content
• The duration and uploadDate
• A transcript if available -- this is the single most impactful property for AI citability

Common Implementation Mistakes

1. Missing thumbnail -- Google requires thumbnailUrl for video rich results. Without it, the Schema is incomplete. Use a high-quality, descriptive thumbnail.

2. Generic descriptions -- "Watch this great video about SEO" provides no value. Write descriptions that summarize the video's actual content with specific details.

3. Wrong duration format -- Use ISO 8601 format (PT10M32S), not "10:32" or "632 seconds."

4. Not including transcript -- The transcript is the most valuable property for AI because it makes the video's spoken content searchable and citable. Invest in transcription.

5. Duplicate VideoObject on multiple pages -- Each video should have VideoObject Schema on one canonical page only. If the same video appears on multiple pages, add Schema only to the primary page.

6. Stale view counts -- If you include interactionStatistic with view counts, keep them updated. Stale numbers reduce credibility.

Validate your implementation with our testing structured data guide.