ZTOA Group繁
ZTOA Group

Insights/Does video help AI search? What the data and Google's docs say

Does video help AI search? What the data and Google's docs say

By Kenny Tang · 2026-10-23 · 5 min read

Does video help AI search? What the data and Google's docs say

In short

  • Video can help, mainly through YouTube. A BrightEdge analysis of AI answers from May 2024 to September 2025 found YouTube was the most-cited domain in Google AI Overviews and AI Mode, with an average citation share of about 20% across the AI platforms it tracked.
  • AI tools mostly read words, not pictures. Titles, descriptions, captions and transcripts are what make a video findable and quotable.
  • Video is not a requirement. Google says there is nothing extra to do to appear in its AI features; good video supports good written content, it does not replace it.

What the data shows

BrightEdge, a search analytics company, tracked how often YouTube was cited across Google AI Overviews, Google AI Mode, Perplexity and ChatGPT between May 2024 and September 2025. Its headline findings:

  • Google AI Overviews: YouTube had a 29.5% citation share and was the most-cited domain in its data.
  • Google AI Mode: a 16.6% share, again the most-cited domain.
  • Perplexity: a 9.7% share, fifth among cited domains.
  • ChatGPT: only 0.2%, although that figure was rising from a very small base.

YouTube was cited far more than any other video platform: BrightEdge put Vimeo at 0.1%. The numbers moved a lot from week to week, so treat them as a direction, not a fixed rate. The pattern is clear enough, though. When Google's AI tools reach for a video, they usually reach for YouTube, while ChatGPT rarely does.

BrightEdge also looked at the kind of question. Videos were cited more for tutorials on a specific platform or product, demonstrations of tools, product reviews and comparisons, and questions where seeing something done helps. They were cited less for strategy questions, career advice and abstract topics with no visual benefit.

What Google says

Google's guidance on AI features is short. There are no additional requirements to appear in AI Overviews or AI Mode, and no special optimisation is needed. A page must be indexed and eligible to show a snippet in normal Search, and even then inclusion is not promised. Among general good practice, Google lists supporting your written content with high-quality images and videos where they help.

For video itself, Google's video SEO guide sets out the basics:

  • A dedicated watch page for each video, where the video is the main reason to visit, with a title and description unique to that video. A blog post that simply embeds a video alongside other content usually does not count as a watch page.
  • A stable thumbnail that Googlebot can fetch, at a URL that does not change.
  • Crawlable video files. Google needs to fetch the file to show previews and key moments, so the file should not be blocked by robots.txt or noindex.
  • VideoObject structured data. Google's required properties are name, thumbnailUrl and uploadDate; recommended ones include description, duration, contentUrl and embedUrl.
  • Key moments. Clip markup lets you mark chapters with start times and labels, so Google can send people to the right part of a long video.

Schema.org also defines a transcript property for VideoObject, described as the transcript of the video. Google's video markup page does not list it, so add it for other tools and for completeness, not as a Google ranking factor.

Why words matter more than pictures

An AI assistant writing an answer works mostly with text. For a video, that text is the title, the description, the chapter labels, the captions and any transcript published on the page. A ten-minute video with a vague title and no captions gives an AI tool very little to quote. The same video with a clear title, a summary, chapters and an accurate transcript gives it plenty.

This is where Hong Kong businesses often lose out. Many local videos are recorded in Cantonese with English or Chinese on-screen text but no captions. YouTube does create automatic captions in Cantonese/Hong Kong and in Chinese, and it also warns that machine captions vary in quality and should always be reviewed. Brand names, district names, product codes and mixed Cantonese-English speech are exactly where automatic captions go wrong. Ten minutes of correction makes the transcript something a tool can quote accurately.

Captions also help people. Many viewers watch on a phone with the sound off, and a transcript lets them read the key points in seconds.

Five steps for Hong Kong brands

  1. Start with questions, not with video. Pick topics where showing helps: how a product is installed, what a space looks like before and after, how a service works step by step. BrightEdge's data suggests these are the videos most often cited.
  2. Publish on YouTube and on your site. Upload to your YouTube channel, then give each important video its own watch page on your website with a unique title, a written summary and the embedded player.
  3. Write the words around the video. Use a clear, specific title, a description that answers the question in the first two sentences, and chapter labels for anything longer than a few minutes.
  4. Fix the captions and publish the transcript. Review YouTube's automatic captions in Cantonese or English, correct names and terms, then add a cleaned-up transcript below the video on your watch page. Do it in both English and Traditional Chinese where your customers use both.
  5. Add the markup and check access. Add VideoObject structured data with name, thumbnailUrl, uploadDate, description and duration, add Clip markup for chapters, and make sure robots.txt does not block your thumbnails or video files.

If you are already publishing case studies, a short walkthrough video on the same page is a natural place to begin; see our portfolio for the kind of projects that suit it. Our AI marketing and GEO service covers video pages as part of an AI-visibility plan.

The short version

Video helps AI search when it comes with words. YouTube is cited often in Google's AI answers, but what an AI tool actually reads is the title, the description, the chapters and the transcript. Make one good watch page per video, correct the captions in the languages your customers use, and add the markup Google documents. If you want to know where video would make the biggest difference for your brand, contact us for an AI-visibility check.

Frequently Asked Questions

General

Do I need video to appear in Google AI Overviews?+

No. Google says there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode; a page must be indexed and eligible to show a snippet. Google lists high-quality images and videos as support for your written content, not as a requirement.

Should we host videos on YouTube or on our own website?+

Usually both. A YouTube channel gives your videos a place on a platform that AI tools cite often, while a watch page on your own site, with its own title, description and markup, keeps the content and the enquiry on your domain. Google recommends one dedicated watch page per video.

Does YouTube create captions in Cantonese?+

Yes. YouTube's help centre lists Cantonese/Hong Kong and Chinese among the languages for automatic captions. YouTube also says automatic caption quality may vary and that you should always review and correct them, which matters for brand names, place names and technical terms.

Sources

Ready to make your brand the focus?

Tell us about your project and we will get back to you.