YouTube has long been the platform people go to learn how to do things. Now it seems AI systems are increasingly sending them there too.
Across LLMs and AI search products, YouTube is emerging as a major source of citations. Digiday has broken down exactly how it got here.
YouTube continues to be cited across LLMs #
YouTube’s growing role in AI search isn’t being driven by a single model. Meltwater’s data showed that YouTube was the most-cited platform by eight AI platforms (Claude, ChatGPT, Copilot, Google AI Mode, Google AI Overviews, Gemini, Perplexity and Grok) in August, while it continues to surpass Reddit, which used to hold the lead after Google struck a deal with the platform to make its content available for AI products. The shift suggests YouTube is becoming more than a video destination and increasingly part of the information layer that AI systems use and direct users toward.
Lauren Lyster, vp and head of social media at Go Fish Digital, said she believes YouTube’s advantage is partly due to how easily its content can be surfaced and processed by AI systems.
“YouTube is obviously where you can work with creators, and that is going to be content that gets transcribed and ranked by services,” she said.
YouTube is (unsurprisingly) most cited in Google models #
YouTube is becoming more prominent within Google’s own AI search products, AI Mode, AI Overviews and Gemini. According to Tinuiti’s Q2 2026 AI Citation Trends report, YouTube’s share of citations in Google AI Mode more than quadrupled between January and April 2026, while it’s share in AI Overviews more than doubled.
The data comes from Tinuiti’s analysis with AI visibility platform Profound, which tracks citations generated from a consistent set of commercial-intent prompts across AI platforms. It shows YouTube content is increasingly being pulled into Google’s AI-generated search experiences. But a citation doesn’t necessarily mean YouTube content was used to generate the answer itself.
“YouTube, especially in relation to Gemini, is a big priority for our clients because its answer engine isn’t just crawling the videos but the context of those videos and everything else that goes into producing them,” said Crystal Duncan, evp of brand engagement at Tinuiti.
YouTube’s tutorial content is cited, even when it isn’t specifically requested #
YouTube is generally considered the dominant platform for “how-to” content, so it’s unsurprising that YouTube videos feature prominently in LLM responses.
What’s more interesting, according to Weglot, is what those citations actually represent. When comparing YouTube citations in Claude and Gemini, both models frequently embed YouTube videos as recommended content. The citation doesn’t necessarily mean the model drew on YouTube to generate its answer — it can simply be pointing users to a video as a source for further information.
Longer videos get cited more than Shorts #
Popularity alone doesn’t fully explain which YouTube videos get cited. Data from OtterlyAI, that was cited on DEV Community (an online developer community) by a user called Watson Foglift, points to the structure within the video itself, beyond just the words.
Long-form content seems to dominate as 94% of YouTube citations go to long-form videos, compared to just 5.7% which Shorts account for. And the sweet spot seems to be YouTube videos that last between 10 and 20 minutes, which account for 32.1% of citations.
One possible explanation is that videos of this length are more likely to include things like chapter markers and timestamps, which gives LLMs granular signals about the content and where specific topics appear. Descriptive, keyword-aligned titles and descriptions can then provide additional context about what the video covers.
Think of it in the same way a human searches. When looking for something specific in a book or long document, you’re more likely to head straight to the relevant chapter or section than read the whole thing. These same structural markers can make a wealth of video content easier for AI systems to understand and reference.
Deloitte Digital’s head of content, creator and AI, Jenny Kelly, is seeing a similar shift in how brands approach video, with content increasingly structured in ways that make it easier for AI systems to discover and understand. That can mean using more natural language, incorporating studies and adopting FAQ-style formats, she explained.
“They are literally setting the content stage for what the LLMs are looking for and how they prioritize what it is,” Kelly said.
YouTube‘s citations don’t necessarily last #
YouTube may be becoming a major source for AI search, but those citations can be short-lived.
Data from Scrunch, an agent experience platform**,** found that YouTube citations faded by around 43% within a month, with the citation rate falling from 18.4% to 10.5%. The data, which tracked videos for the month after they were first cited and measured how often they continued to appear in AI answers, came from Scrunch’s analysis of thousands of sponsored YouTube videos and millions of citation events across AI platforms between May and July 2026.
The finding seems to suggest that appearing in an AI answer isn’t necessarily a lasting win for creators or brands. AI systems are constantly refreshing their retrieval and ranking systems, meaning newer videos can displace older ones and the topics being searched can shift over time.
More in Marketing
[Marketers have bought into microdramas — but a lack of ROI might make them irrelevant](https://digiday.com/marketing/r-i-p-quibi-you-would-have-loved-microdramas/)
Microdramas are attracting brand dollars and audiences. The harder question is whether the format can translate that attention into meas.
[Some marketers see ‘no ceiling’ for beauty brands advertising in sports](https://digiday.com/marketing/some-marketers-see-no-ceiling-for-beauty-brands-advertising-in-sports/)
By now, many in the beauty space have wised up to just how powerful a tool athletes and sports are for their brands.
[The case for and against retail media networks as brand-building channels](https://digiday.com/marketing/the-case-for-and-against-retail-media-networks-as-brand-building-channels/)
Retail media networks pitch brand-building on shopper data, but ROI metrics, costs and budget fights hold them back.