Videos in AI Overviews: why YouTube dominates and how your B2B brand gets cited

Anyone asking Google a question today increasingly gets no list of blue links, but a summarized answer at the top: the AI Overview. And in that answer a video appears strikingly often. Not as decoration, but as a cited source on which the model bases its answer. For B2B marketers who thought in terms of text and keywords for years, that is a fundamental shift. The question is no longer only whether your page ranks, but whether your content is fit to be cited by a generative system.

The numbers leave little room for doubt. According to research by BrightEdge, YouTube is the most-cited domain in Google’s AI Overviews, and no other video platform comes close. That has everything to do with how AI understands video and with which video formats are ready to be cited. In this article we explain why AI surfaces video, why YouTube is so dominant, what role transcripts and structured data play, and how you, as a B2B brand, capitalize on videos in AI Overviews in concrete terms.

Test your AI visibility: score your page with our GEO check.

Why AI specifically surfaces video

An AI Overview wants to answer a question as completely as possible with the most reliable and best-structured source. For an entire category of queries, video is simply the strongest answer format. For procedural questions (“how do I set up X”, “how does process Y work”), demonstrations of software or physical products, and visual comparisons, a video delivers more value than a wall of text. The model recognizes this and presents the video as a logical supplement to, or replacement of, a text answer.

At the same time, video is only usable for a language model once it can read the content. An AI does not literally look at the image the way a human does, it processes the textual representation of that video: the title, the description, the chapters and above all the transcript. A well-captioned, clearly structured video is therefore in effect a text document with visual support, precisely the kind of source a model likes to cite.

A striking detail from a large-scale study by OtterlyAI, based on more than 100 million citation instances, is that popularity metrics such as views have virtually no bearing on the likelihood of a citation (a correlation of roughly -0.03). AI does not select the most-watched video, but the most usable and best-structured one. That is good news for B2B: you do not need to build viral channels, you mainly need to be citable.

YouTube as the dominant source: what the numbers show

The numbers on video citations are downright one-sided. YouTube is not merely the biggest video player in AI answers, it is practically the only one that matters. The table below summarizes the key findings from BrightEdge’s analysis, published via Search Engine Land.

The message is clear: other video platforms are in practice invisible in AI search results. Whoever wants to be cited via video publishes on YouTube. That is no coincidence. YouTube belongs to Google, is thoroughly indexed, automatically supplies transcripts and supports chapters and rich metadata, all signals that make it easy for a model to understand and cite the content.

A second large-scale study nuances the picture further: within social platforms specifically, YouTube ranks second behind Reddit, with about a third of all citations from social sources. But across all domain types, YouTube stays the front-runner. For B2B brands thinking about their presence in AI answers, that is a clear signal to treat YouTube as a strategic channel, not as a side matter. Read more about the fundamentals in our guide on YouTube SEO.

The role of transcripts and structured data

If AI “reads” video via text, then the transcript is your most important lever. A missing or sloppy automatic transcript means the model has to guess at your content. An accurate, complete transcript, by contrast, gives the model exactly the sentences it can lift into its answer. Treat your transcript therefore as content in its own right: correct errors, add technical terms correctly, and make sure important statements are phrased literally and clearly.

For videos on your own website, structured data comes into play on top of that. Google advises in its official documentation on video to use VideoObject markup, with properties such as a unique thumbnail, the contentUrl, and above all Clip and SeekToAction markup to explicitly name chapters and timestamps. With that you make crystal clear to the machine which segment answers which sub-question. Dive into the details via our article on schema markup.

The combination is what counts. A video with a clean transcript, descriptive chapters and correct VideoObject markup offers an AI system three mutually reinforcing signals: what the video covers, how it is built up and which fragment is relevant for which question. That is exactly the kind of reference value that gets preferred over standalone engagement figures.

A practical approach for B2B brands

Presence in AI Overviews via video is not a matter of churning out content en masse, but of publishing in a targeted and citable way. A workable approach for B2B looks like this:

1. Start with the questions, not with the camera. Inventory the concrete, procedural and comparative questions your audience asks in the consideration phase. These are the questions where AI most often places a video alongside or within its answer.
2. Make one video per clear question. Avoid long, all-encompassing videos. A targeted video of a few minutes that fully answers one question is easier to cite than an hour-long webinar.
3. Structure with chapters. Give each chapter a title that literally names the sub-question. That way a model can cite exactly that segment.
4. Deliver a correct transcript. Check the automatic captioning and rewrite where necessary. Also place the core as text in the description.
5. Publish on YouTube and embed on your site. Combine YouTube’s reach with VideoObject markup on your own page, so both channels feed the AI.
6. Measure your AI visibility. Track whether and how your videos get cited, and adjust for the questions where you do not yet appear.

This approach dovetails seamlessly with a broader video strategy. For optimizing the findability and relevance of your videos, we refer you to our guide on video SEO. And whoever wants to grasp the bigger picture, namely how you become the source material that AI draws from in the first place, is best off reading our article on becoming a source for AI.

Conclusion: video is no longer a side matter

The shift to AI-driven search results turns video into a serious channel for B2B visibility. Not because video is trendy, but because AI systems demonstrably prefer video for a growing set of questions, and because YouTube is by far the most-cited domain. The good news: because citations revolve around structure and reference value rather than viral figures, specialized B2B brands have a fair chance. Whoever builds their videos around concrete questions, equips them with clean transcripts and correct markup, and consistently publishes on YouTube, step by step increases the chance of appearing in the answer that millions of users get to see.

Frequently asked questions

Need help?

Want to translate this into execution? See how we approach this with AI visibility.