Research · YouTube in AI search · October 2026
How important is YouTube for AI search?
To find out, we analysed 197,314 AI answers to more than 1,600 real buying and research questions, tracked daily across five leading AI engines (ChatGPT, Google AI Overviews, Gemini, Perplexity and Copilot) over eight weeks in the UK and US. Then we studied the 2,800 YouTube videos those answers cited: how many views they had, how big their channels were, how old they were and how long. Google AI Overviews cited a YouTube video in almost one answer in three. The other four engines almost never did.
of Google AI Overview answers cited at least one YouTube video. For ChatGPT it was under 1%.
of the videos AI cited had fewer than 10,000 views. Only 1% had a million or more.
came from channels with fewer than 10,000 subscribers.
of YouTube citations went to just one in ten of the videos AI cited.
01 · By engine
Which AI engines cite YouTube? Only Google AI Overviews, in practice
Google AI Overviews cited a YouTube video in almost one answer in three. ChatGPT cited one in under 1% of answers, and in seven of our ten categories it never did. Gemini, which Google also owns, cited YouTube in 0.1% of answers. Perplexity and Copilot almost never did.
Share of answers citing at least one YouTube video
Each category weighted equally. Google AI Overviews and ChatGPT tracked in ten categories, Gemini in five. * Tracked in two categories only, so treat as a guide.
Google AI Overviews cited YouTube in 31% of answers and Reddit in 19%, measured the same way over the same eight weeks. Read the Reddit study.
of all the sources Google AI Overviews cited were YouTube videos: one source in every ten, across every category we track.
02 · By category
How much YouTube matters depends on your category
The 31% is an average across very different markets. Questions about learning a new skill pulled in a video in four AI Overviews out of five. Questions about hotels almost never did. Anything people would rather watch than read, from software tutorials to car reviews, leans heavily on YouTube.
Share of Google AI Overview answers citing YouTube
Each category is one set of questions tracked in reconnAI. Personal finance is UK only.
03 · Views and subscribers
Do you need a popular video? No.
You might expect AI to pick the videos everyone else is watching. It doesn’t. Nearly three quarters of the videos it cited had fewer than 10,000 views, and almost four in ten had fewer than 1,000. AI is choosing videos that answer the question, not videos that are popular.
View counts of the videos AI cited
of cited videos came from channels with fewer than 10,000 subscribers. Only 23% came from channels with 100,000 or more.
of all YouTube citations went to videos with fewer than 1,000 views. A video hardly anyone has watched can be quoted to thousands of people through AI.
A real example
244 views on YouTube. 651 AI answers.
In one category, the most-cited video of the whole study had been watched 244 times. Google AI Overviews cited it in 651 answers to 58 different questions, on 51 of the 56 days we tracked. It is a 12-minute walkthrough from a channel with about 3,000 subscribers, posted ten months before our study began. Google cited it in more of our answers than it had views on YouTube.
View and subscriber counts are as recorded by reconnAI, so a video that AI cited may have gained views since. Based on the six categories with more than 100 cited videos, each weighted equally. Rounded to whole numbers.
04 · A short list of videos
Which videos does AI cite? The same few, again and again
AI doesn’t spread its YouTube citations evenly. On average, the top 10% of cited videos in a category took 60% of all its YouTube citations, and the top 10% of channels took two thirds. Once AI settles on a video, it keeps coming back to it.
Where the YouTube citations went
of YouTube citations went to the top 10% of channels in each category. In some categories, a single channel took one citation in four.
05 · Length and age
Long videos, not Shorts
Shorts dominate YouTube’s own feed, but not AI answers. Only 8% of the videos AI cited were Shorts. Four in ten ran for 10 minutes or more. And unlike Reddit, where most cited threads were more than a year old, almost half the cited videos were less than a year old.
How long the cited videos were
Videos of 10 minutes or more took 41% of all YouTube citations. Shorts took 7%.
How old the cited videos were when AI cited them
47% of cited videos were under a year old, against 32% of the Reddit threads AI cited. New videos get picked up faster than new threads.
Age is the time between a video being posted and the first answer in our study that cited it. Based on the six categories with more than 100 cited videos, each weighted equally. Rounded to whole numbers.
06 · What it means
A YouTube strategy for AI search
If your buyers ask Google questions, YouTube videos are already part of the answers they get. Four things to take from the data.
It’s a Google story
YouTube’s influence on AI answers runs through Google AI Overviews. If your buyers use Google, video matters. If they only use ChatGPT, it barely does.
You don’t need a hit
Most cited videos have under 10,000 views and come from small channels. A clear video that answers one real question can be cited hundreds of times.
Know whose videos AI uses
A short list of videos and channels shapes your category. Find them, watch them, and know what they say about you and your competitors.
Go long, not short
Shorts were 8% of cited videos. Walkthroughs, reviews and tutorials of several minutes are what AI cites, and new ones get picked up within months.
How reconnAI measures this
See the videos behind the answers
This study is built on the same daily data reconnAI collects for every customer, including every source each AI answer cites.
YouTube’s role depends on the engine
Every engine, every day. Your buyers’ questions run daily on the AI engines that matter to you, with every source each answer cites, so you can see where video shows up and where it doesn’t.
A few videos take most of the citations
The videos, ranked. See which YouTube videos and channels AI cites for your questions, and how often each one comes up.
Views don’t decide what gets cited
Each video’s details. See a video’s views, channel size, length and age alongside how often AI cites it, so you can spot the quiet videos doing the work.
Every category is different
Your category, not an average. See how much YouTube matters for your questions, alongside Reddit and every other website AI cites.
Questions
Common questions
Does YouTube influence AI search results?
Yes, through Google. In our study, Google AI Overviews cited a YouTube video in 31% of answers, more often than they cited Reddit. ChatGPT, Gemini, Perplexity and Copilot each cited YouTube in under 1% of answers. It also depends on your category: YouTube appeared in 82% of AI Overviews in one category and under 1% in another.
Can YouTube videos get featured in Google AI Overviews?
Yes, and often. In the six categories we analysed in detail, Google AI Overviews cited 2,800 different videos in eight weeks. One video in ten took 60% of those citations, and most of the videos cited had fewer than 10,000 views.
Does ChatGPT cite YouTube videos?
Rarely. ChatGPT cited a YouTube video in 0.6% of answers in our study, and in seven of our ten categories it never did. This measures citations, meaning the sources linked with an answer. It doesn’t tell us what an AI model learned from when it was built.
Do you need lots of views for AI to cite a video?
No. 72% of the videos AI cited had fewer than 10,000 views and 38% had fewer than 1,000. The most-cited video in one category had 244 views and appeared in 651 AI answers. 42% of cited videos came from channels with fewer than 10,000 subscribers.
Does AI cite YouTube Shorts?
Not much. Shorts made up 8% of the videos AI cited and took 7% of YouTube citations. Videos of 10 minutes or more were 39% of cited videos and took 41% of citations.
How do I optimise YouTube videos for AI search?
Start by finding the videos AI already cites for your buyers’ questions and see what they have in common. In our data, the videos AI cites are usually several minutes long, answer one specific question, and do not need many views. We measured which videos get cited, not what makes a video get cited, so track any new video over time to see whether AI picks it up.
Method
How we measured it
reconnAI puts the same real buying and research questions to AI engines every day. For this study we used more than 1,600 questions from ten categories, asked in the UK and US on the free version of each engine, between 31 August and 25 October 2026. We left out questions that name the company being tracked.
We counted an answer as citing YouTube when at least one of its sources was a youtube.com or youtu.be link to a video. For the video analysis we used the six categories with more than 100 cited videos, 2,832 videos from 1,605 channels in all, and matched each one to its view count, channel subscriber count, posting date and length as recorded by reconnAI. We split the eight weeks into two four-week months to see whether AI kept citing the same videos.
Every figure is worked out for each category first and then averaged, so each category counts equally and no single one dominates. Because YouTube’s role varies so much by category, the headline 31% has a wide range: we are 95% confident the true average is between 19% and 46%. All data is anonymised.
Which videos is AI quoting about you?
We run your buyers’ questions every day across the major AI engines and show you the sources behind every answer, YouTube videos included.