According to new research from Jellyfish, creators are becoming vital sources for AI chatbots. The digital marketing agency found that more than 25% of chatbot prompts are answered with responses that reference content from YouTube creators.
Jellyfish, which shared its findings exclusively with Adweek, reported that more than one million unique YouTube videos are cited by AI chatbots in the U.S. each day — and that’s just in the CPG category. Certain “high-intent” subjects like consumer electronics or financial services produce more frequent references to creator videos. In some categories, nearly 50% of responses allude to YouTube creator content.
YouTube’s effort to shore up its referrals has paid off
Once upon a time, Reddit was a more common source than YouTube among AI chatbots. By the start of 2026, however, YouTube had surpassed the “front page of the internet” to become the top referral destination.
Subscribe to get the latest creator news
To pull off that coup, YouTube experimented with features that were pulled from Reddit. A feature dubbed “threaded comments” made YouTube comment sections more conversational, and by extension, more legible for AI models.
YouTube videos are cited by Google’s Gemini model, of course, but Jellyfish also researched responses delivered by other AI chatbots, including Claude, DeepSeek, Meta AI, ChatGPT, and Perplexity. YouTube’s videos show up across all those models, and that’s both a win for Google’s video platform and a source of anxiety for the creators who call that platform home.
What does YouTube’s AI referral boom mean for creator agency?
Creator complaints about unauthorized AI training are almost as old as the chatbots themselves. Tech review king Marques Brownlee was one of the first YouTube stars to call out generative AI services for mimicking his style without his explicit consent. Brownlee is now leading the charge as creators lobby for stronger protections against unlicensed scraping.
YouTube has made it much easier for creators to opt out of AI scraping, but there are still piles of YouTube videos that are being used to train models like Gemini. And when other tech companies with AI ambitions become involved, ethical standards become harder to uphold.
In a class action filing earlier this year, a group of creators argued that companies like Amazon, Apple, and OpenAI bypass YouTube’s rules to access unauthorized AI training materials. Per the complaint, those companies use “scraping tools that are designed to evade and circumvent YouTube’s access restrictions.”
The Jellyfish report shows that creators are right to be concerned about tech companies raiding YouTube to train their AI models. It’s hard to say how many of the channels cited by chatbots gave their informed consent regarding the reuse of their videos, but there is at least one positive creators can take away from the Jellyfish data.
Long-form is back, baby!
Once upon a time, watch time was the most important statistic on YouTube. Long-form videos, and the extensive watch time they deliver, were a major factor in the rise of the swoopy-haired Swede we know as PewDiePie.
The arrival of YouTube Shorts seemed to swing the dial back toward short-form videos, but AI is changing the metagame once again. According to Adweek, AI chatbots prefer videos from “niche creators with videos longer than 10 minutes.” So if your video essay hub or how-to channel has been looking for that extra bit of reach, AI might be able to provide it — whether you like it or not.
