Video SEO: what makes a video get cited by AI?

Almost every team that produces video optimizes for the platform where it publishes and stops thinking beyond that. Video SEO begins exactly after that point: in what Google can crawl, what AI can read, and what the site's own page offers to whoever arrives through search.

In that layer outside the platform, video stops being an asset of one network and becomes a search asset. In 2026, that layer gained weight: YouTube is the most cited domain in AI Overviews, and Search Console started measuring Instagram's, TikTok's and YouTube's performance in search.

Google, for its part, says there's no special optimization to show up in AI features. What exists is the basics done well, and this article shows what those basics are.

What is video SEO, and why does it go beyond the platform?

Video SEO is optimizing the video and the page hosting it so it can be found through organic search, AI answers and each platform's internal search. Optimizing for YouTube or TikTok is one part; the other part is making that same content pay off outside of where it was published.

Caption: Video SEO starts where platform optimization ends: at the page and the structured data.

The distinction matters because the systems read different things. A platform's internal search uses its own signals, like retention and engagement. Google and AI use what they can crawl: text, structured data, a thumbnail, and the page the video lives on. A video can be excellent on the platform and invisible in search.

The data point that changed the math in 2026 is video's weight in AI answers. A 5W survey of 46 million citations found YouTube as the most cited domain in AI Overviews, with 23.3% of citations, ahead of Wikipedia. Video became a source for answers, not just entertainment.

At the same time, Google states that there are no additional requirements or special optimization to appear in AI Overviews or AI Mode. The practical takeaway is that video SEO in 2026 is ordinary SEO applied to video with rigor, and that's what the following sections break down.

How does Google index videos?

Google indexes a video when it finds a public page dedicated to it, with its own title and description, the video visible as the main content, an accessible thumbnail, and, ideally, structured data describing the file. Without these elements, a video can exist on the page and never show up in video results.

The official documentation is specific about the page. According to Google, each video should have its own display page, with a unique title and description, and that page needs to be indexed and performing well in search before the video gets considered. The video also can't stay hidden behind other page elements.

Since December 2023, Google's "Videos" tab only shows pages where the video is the main content. A long article with a video embedded in the middle doesn't compete there; a page built around the video does. It's an architecture decision, not an editorial one.

The thumbnail has its own rules. Google accepts formats like JPEG, PNG, WebP, SVG and AVIF, requires a minimum size of 60 by 30 pixels (bigger is better), and asks for a stable, unique URL per video. A generic thumbnail reused across multiple videos confuses the crawler and weakens indexing.

The video sitemap rounds out the set. It tells Google where the file or player is, the thumbnail, the title, the description and any regional restrictions, and speeds up discovery on sites with many videos. On small sites, the structured data on the page itself is usually enough.

How does AI pick which video to cite?

AI picks the video it cites based on the text it can read about it, not popularity. The Otterly.ai study of more than 100 million citations found that 40.83% of cited videos have fewer than a thousand views, and that the correlation between views and citation is practically zero.

What matters, per the same study, is the description. Long descriptions had a positive correlation with citation (an average of 334 words among cited videos), and 31% of citations pointed to a timestamp on a specific segment. That indicates AI is reading chapters and transcripts to locate the answer inside the video.

Format also decides. 94% of video citations go to long-form videos, and only 5.7% to Shorts. Short video serves reach in the feed; long-form video, with structure and supporting text, serves citation. They're two video production investments with different returns.

Each tool has a different appetite for video. In the Otterly.ai survey, YouTube accounted for 38.7% of citations on Perplexity and 36.6% on AI Overviews, but only 4.4% on ChatGPT and 0.2% on Gemini. Whoever depends on Google-driven traffic has more to gain from video than whoever targets chat assistants.

The lesson for scripting is to write for two readers. The person watches; AI reads the title, the description, the chapters and the transcript. If the answer to the question that names the video isn't written down in any of those places, it doesn't exist for the machine.

How do you optimize your video's page on your own site?

A video's page on your site is optimized with VideoObject structured data, a block of text summarizing the content, time-stamped chapters, a visible transcript and its own thumbnail. It's the version of the video that Google and AI read in full, and it's what drives organic traffic to the brand's own domain, not the platform.

VideoObject has three required properties in Google's documentation: name, thumbnail URL and upload date. Recommended ones include the content URL (preferred over the player URL), description, duration and the list of clips. Without the required ones, the markup is ignored.

Key moments are the most underused feature. With the Clip property, each segment gets a name, a start time and its own URL, and Google can display them as direct access points into the video. The documentation requires a minimum duration of 30 seconds per segment and lets the platform read the markers automatically when the site has a consistent structure.

The transcript closes out the set. Published as text on the page itself, it hands AI exactly what the video says, with the keywords the audience uses when asking. It also serves people watching without sound: a Verizon Media study with Publicis found that 69% of people watch video without audio in public places.

The same logic as internal links applies to a video page. It needs to receive links from related articles and link back to them, or it's born orphaned. A video with no textual context around it is a file, not a page.

What changes in video SEO on social media?

Video SEO on social media changed in 2025 and 2026 because Google started indexing and measuring public content from Instagram, TikTok, X and YouTube in its own search carousels. A video published on a network now competes in the SERP, and the brand can see that performance in Search Console.

The milestone was July 2026. Google launched platform properties in Search Console for Instagram, TikTok, X and YouTube, with click, impression and query reports for posts and videos that show up in search and Discover.

Each account is verified individually, and content shows up in short-video carousels, recent-posts carousels, and "what people are saying" panels.

Instagram had already opened the door. Since July 2025, public content from professional accounts of users over 18 can be indexed by search engines, with the option on by default. Reels and carousels gained a life outside the app, and the caption became SEO text.

The indexing of social content connects video SEO to SEO for social media. The video's title, caption and hashtags are read by Google as page text, and the search-intent rule applies just the same. What's new is that it can now be measured, as detailed in Search Console for social media.

On YouTube specifically, chapters and description are what AI reads. Manual chapters require the first marker at 00:00, at least three markers in order, and ten seconds per chapter. The full rule for optimizing inside the platform is in SEO for YouTube.

How do you measure video SEO?

Measuring video SEO requires looking at three places: the site's Search Console (for your own video pages), Search Console's platform properties (for YouTube, Instagram and TikTok), and AI citation monitoring tools. None of them alone shows the whole picture.

In the site's Search Console, traffic coming from AI Overviews and AI Mode shows up inside the "Web" filter, with no separation, according to Google's documentation on AI features. What can be isolated is video-page performance through the "Video" search-type filter and the VideoObject rich-result reports.

In the platform properties, the reading is by content: which videos gained impressions and clicks in search, for which queries and in which carousel. It's the first time Google returns this data for social content, and it points to which videos are worth repurposing on the site as their own page.

AI citation still has no official dashboard. Third-party tools monitor which URLs show up in answers, and a brand can run a manual test with the questions that name its videos. It all organizes into a simple table:

Table: Video SEO metrics and where to find them, based on 2026 Google Search Console documentation.

It's the same rule that applies to any page in Google's AI Mode: impressions without clicks grow, and value shifts to citation and brand recall, not just visits.

How does education marketing apply video SEO?

In education marketing, video SEO turns open classes, testimonials and course videos into the institution's own site pages, with transcript and structured data, so they show up when a prospective student searches the career or the question. It's one sector example: the logic serves any brand that produces educational video about its own product.

In student recruitment, the question a prospective student asks Google or AI is rarely the course's name. It's "is it worth studying software engineering" or "how much does a physical therapist earn."

A video with a coordinator answering those questions, published on its own page with a transcript, competes in organic search and AI Overviews for the right term, the way video content for recruitment already does.

Repurposing is the secret to cost. The same video goes to YouTube with chapters and a long description, becomes a Reel with an optimized caption, and gets a site page with VideoObject. Three search surfaces from one production, and the site page is the only one that brings the lead into the institution's CRM.

What doesn't work is embedding a YouTube video at the end of an article and expecting it to rank as video. Under the 2023 rule, that video isn't the page's main content and doesn't show up in the videos tab. It helps the article; it doesn't earn a life of its own.

Frequently asked questions about video SEO

What should you prioritize first in video SEO?

The first priority in video SEO is your own page: an address on your site for each relevant video, with a complete VideoObject, a visible transcript and marked chapters. It's the only place where the video works for your domain and your CRM, and it's what Google requires to treat it as main content.

The second priority is a long description and chapters on YouTube, because that's what AI reads to cite. The third is verifying social accounts in Search Console's platform properties and finding out which videos are already paying off in search without anyone having measured it. Views remain a platform metric; citation and traffic are search metrics.

Video SEO is one piece of a bigger picture, in which the answer arrives before the click and citation is worth more than position. To understand that picture and its effect on organic traffic, the next step is what changes with Google's AI Mode, where the same principles apply to text and video.