AI & innovation, SEO Strategy 16.12.25

AI isn’t watching your videos: Schema Markup for video discoverability

AI isn't watching your videos: Schema Markup for video discoverability

This video provides a concise technical guide on optimising video content for AI search engines, emphasising that AI "reads" the accompanying page rather than "watching" the video. Varn CEO Tom Vaughton outlines three crucial steps using JSON-LD structured data (specifically VideoObject schema) and written context to improve discoverability in rich results and AI answers.

Download Tom's Video Schema Cheat Sheet

Structured data is how search engines and AI systems actually read and understand your videos. Our ‘cheat sheet’ gives you a clear, practical 6-point framework to ensure every video you publish is fully optimised for AI-driven visibility.

Download the cheat sheet

Search is shifting from type and scroll, to conversational answers, and video is an important part of this if you want to keep up.

No, AI doesn’t watch videos. AI crawlers read your page and the code behind it. That means the visible text, a clear transcript, and structured data are essential in getting your videos discovered in AI search.

AI bots and search crawlers read your page, and the code behind your page. That means the visible text, a clear transcript, and structured data that you provide behind the scenes are essentioal for being understood, indexed and surfaced in AI answers and rich results.

Start with the VideoObject schema in JSON. As a minimum, you’ll need the name, thumbnailURL, and the uploadDate.

Further strong signals to include are description, duration, and either the contentUrl (video file) or the embedURL (the video player). Ideally, add both. For even further context, include inLanguage, some keywords, the about or mentions for the entities, the publisher,  the author, and a last review date.

For more information on VideoObject schema, plus a paste-ready JSON-LD block, download our free Video Schema Cheat Sheet.

Make key moments clear is essential. AI and Google prefer videos that you can deep link into, so label your YouTube video chapters with clear answer-driven labels that make sense. In your companion page, use the SeekToAction schema so that your search results can jump straight to the answer. Alternatively, add Clip items with a name, startOffset and endOffset.

Video transcripts and captions are essential for AI and search bots to discover your video. Adding a clean transcript on the page and accurate captions on the video provides textual context which the AI uses to understand and surface your content. This improves understanding, accessibility, and your chances of being quoted.

Make sure to run some quick validation tests. Before you publish, run the Rich Results Test. Make sure the video bytes are fetchable, don’t block the CDM path, and double check that title, duration, and description match across the page content, schema, and YouTube video details.

Transcript

Introduction

(00:00) I recently gave a quick overview of video and SEO, including that classic video sort of Vimeo versus YouTube debate. Following this, we had a lot of questions about the schema side of things and what they are and how to use them. I’ve recorded this follow-up with three main sort of technical tips and put together a video schema cheat sheet for AI search, including a simple quick six-point checklist you can download and use with your dev team or content team.

Is video important for AI search visibility?

(00:28) As I’ve mentioned before, search is shifting from type and scroll, the traditional, to conversational answers, and video is an important part of this if you want to keep up.

Does AI watch your videos?

(00:38) The challenge is that AI does not actually watch your video or not currently. It reads your page and the code behind your page. That means the visible text, a clear transcript, and structured data that you provide behind the scenes are what matter if you want to appear in AI Answers and those rich results.

The key to video discoverability: JSON-LD schema

(00:56) You need to use what is called JSON-LD structured data on the page to describe the video and the schema vocabulary. For example, video object is one. There’s a description, you have this JSON-LD, it’s in the data, it’s in a tag in the code of your page, but it’s not JavaScript. When you add accurate JSON plus a transcript, you’re actually speaking the machine’s language, the AI language, so that they can understand you and they can cite and deep link to your content. There’s more about this in the downloadable guide mentioned below if you want a bit more context. All that said, and this will be part of the downloadable guide, the three main technical tips you need to follow now to optimise your video and website for AI search in Google are:

How to build a complete VideoObject

(01:42) One is to create a complete VideoObject. So you need to start with the VideoObject in JSON, and as a minimum, you need the name, the thumbnailURL, and the uploadDate. Then you want to add what I class as strong signals. So there are description, the duration, and either the contentURL, which is the actual video file, or the embedURL, which is the player. Ideally, though, you’d do both. For extra context, if possible, the bots also love, in language, some keywords, the about or mentions for the entities, and the publisher, and the author, with a sort of visible last review date. You can keep this consistent with your YouTube metadata, If you use one or a video sitemap and use that as well, it will really help.

Define key moments & deep linking with SeekToAction

(02:27) Tip two is make those key moments really obvious. AI and Google prefer videos that you can deep link into. So if your player supports time parameters, use the seek to action schema so that your search results can jump straight to the answer or add clip items with a title and a start and a sort of end offset. Mirror those segments as YouTube chapters with clear answer-driven labels, not part one, two, three, actually names that make sense.

Why transcripts & captions are essential for AI

(02:52) And three, give it text to read because AI reads. You need to add a clean transcript on the page and accurate captions. Also, fix names and terminology across everywhere. This improves understanding, accessibility, and your chances of being quoted.

Final checklist: Validate your schema & fix common errors

(03:06) Then, when you’re ready, run some quick validation tests. So before you publish, make sure you run the rich results test and sort of get to know what I call blocking errors. Make sure also that the video bytes are fetchable, don’t block the CDM path, and double check that title duration and description that they match across the page, the schema, and YouTube. Again, all of this is linked in the document below. Two issues we see a lot though are actually very quick to fix. They’re the missing transcripts and people only adding the embed URL and hiding the real file. This hurts previews and key moments, so we don’t recommend it. Remember that AI can’t see your answers without text. Finally, if you’d like the copy and paste ready JSON to go in your website, again, download the link below. It’s got the six point schema checklist and literally you can copy and paste some of it in. This is your chance to help the bots help you and your business. So please get up to speed.

Tom
16.12.25 Article by: Tom, CEO More articles by Tom

Do you need help future-proofing your SEO strategy?

Get in touch
cta-background cta-background

Any questions about our blogs?