OtterlyAI’s YouTube GEO Study: What Actually Gets Your Videos Cited by AI Search
YouTube has over 2.7 billion monthly active users and consistently ranks among the top 10 most cited domains in AI-generated search responses. That makes it more than a social media channel. It’s become an essential part of AI Search Optimization.
OtterlyAI analyzed more than +100 million AI citation instances over a 30-day period to understand what drives visibility in AI Search.
Of all citations observed, 5.54% came from social media and video platforms. Within that subset, YouTube represented 31.8% of social media citations.
This study focuses on which YouTube videos get cited in AI Search, not on what drives YouTube channel growth.
Key Findings (TL;DR)
YouTube is the 2nd biggest Social Media source used by AI. YouTube & Reddit make up 78.2% AI citations on Social Media.
Perplexity & Google lead YouTube citation volume across AI platforms. Perplexity (38.7%) and Google AI Overviews (36.6%) drive the majority of YouTube citations, while Gemini (0.2%) and Copilot (0.5%) rarely cite YouTube at all.
94% of AI citations go to long-form YouTube videos, not Shorts Views, likes, & subscribers show near-zero correlation with citation frequency (r ≈ -0.03), indicating AI systems prioritize reference value & structure over popularity signals.
Timestamped YouTube citations appear only within Google’s AI Platforms When a YouTube citation includes a specific timestamp, it shows up in Google AI Overviews (73%) and Google AI Mode (27%). We observed no timestamped YouTube citations in ChatGPT, Gemini, Microsoft Copilot, or Perplexity during the study window.
One structured video can generate multiple AI citations 78% of timestamped videos are cited repeatedly, often across 2–5 chapters, effectively multiplying citation surface area from a single asset.
Why Is Studying YouTube Critical for AI Search Optimization?
AI Search increases zero-click behavior for informational queries because the answer is delivered directly in the interface (for example, Google AI Overviews). Multiple studies have reported measurable CTR declines when AI summaries appear, which raises the stakes for being cited inside the answer. As AI platforms increasingly blend text with videos, YouTube becomes a practical visibility channel for brands trying to stay present inside AI-generated responses.
For brands losing traffic from traditional search, structured long-form video offers a way to stay present inside AI Search rather than outside them.
Scope of Study
This article summarizes OtterlyAI’s YouTube GEO Study: an original analysis of AI citations collected across ChatGPT, Google AI Overviews and AI Mode, Perplexity, Microsoft Copilot, and Gemini. We tracked when AI citations included YouTube video URLs, then analyzed the cited videos’ metadata to identify patterns linked to how often those videos get cited.
AI Search is shifting visibility from “ranking” toward being included as a cited source inside AI generated answers. As AI Overviews and other AI Search Engines generate answers using query fan-out and supporting links, being cited increasingly matters more than simply ranking. At the same time, independent studies show lower click-through rates when AI summaries appear, raising the stakes for AI source inclusion.
This research focused exclusively on YouTube videos that were already cited by AI Search Engines during the observation window.
The central question: What characteristics are associated with being cited more often once a video appears in citation graphs? For each cited video, we collected metadata including:
Format of the video
Video chapter/timestamp inclusion
View count
Likes
Duration
Title word count
Description word count
Channel subscribers
Channel total views
Channel total video count
Publish date and recency
To test linear relationships between citation frequency and individual features, we used the Pearson correlation coefficient, Pearson’s r, which ranges from -1 to +1. Values near zero indicate no linear relationship.
Because this dataset includes only already-cited videos, results are strongest for explaining repeated citation behavior, not initial eligibility.
Social Media’s Role in Generative Engine Optimization (GEO)
Social media plays a measurable but secondary role in AI Search visibility. Based on OtterlyAI’s analysis of 100 million AI citations across six AI Search platforms, approximately 5.54% of citations originate from Social Media and Video domains,
This represents roughly 5.5 million citation instances within the observed dataset. While Brand domains account for the majority of citations at 52.2% and News/Media sources contribute 20.3%, social platforms remain a recurring source category in AI Search. This indicates that community-driven and platform-hosted content contributes to citation ecosystems, particularly for prompts where experiential knowledge, tutorials, or user discussions are relevant.
How YouTube Fits into the AI Search Landscape
YouTube functions as a hybrid source within the Social Media and Video category, combining brand publishing and community content. AI Search engines cite YouTube primarily for demonstrations, walkthroughs, how-to prompts, and comparisons. When videos are aligned to specific prompts and structured with clear titles and metadata, they can expand Domain Coverage and increase citation likelihood across tracked prompt clusters.
YouTube’s Share of Social Media Citations in AI Search
Across 5.5 million social media citations in the dataset, YouTube accounts for 31.8%. Together with Reddit at 46.4%, the two platforms represent 78.2% of all social media citations in AI Search. YouTube ranks as the second most cited social platform, ahead of LinkedIn at 13% and Facebook at 5%, with other platforms each contributing below 3%.
Ranked Social Media Citation Share in AI Search:
Reddit: 46.4%
YouTube: 31.8%
LinkedIn: 13%
Facebook: 5%
Instagram: 2.2%
TikTok: 0.7%
Quora: 0.6%
X: 0.3%
Threads: 0.1%
Vimeo: 0.1%
Social Media Citation Patterns Differ by AI Platform
Social media citation share varies by AI Search engine. Each platform applies different retrieval and citation logic, which changes how often YouTube, Reddit, and other domains appear in AI citations. GEO analysis should therefore be evaluated per platform, not as a single blended average. Let’s explore:
Google AI Mode & AI Overviews: YouTube is king
There seems to be a clear preference in the Google AI ecosystem for YouTube. Over half of Google AI Overviews & AI Mode’s Social Media citations come from YouTube.
Perplexity and ChatGPT Prioritize Reddit
In both ChatGPT and Perplexity, Reddit is the most cited social media domain, accounting for 53.6 to 62.8% of social media citations. LinkedIn ranks second at 25.9%in ChatGPT and 13.9% in Perplexity. YouTube represents a smaller share in ChatGPT at 9.5%, but a higher share in Perplexity at 22.7%.
These differences indicate that social media citation patterns vary meaningfully by platform, with Reddit consistently dominant and YouTube more platform-dependent.
Microsoft Copilot & Gemini: Narrower Social Citation Patterns
Microsoft Copilot differs from all other platforms by prioritizing LinkedIn, which accounts for 43.8% of its social media citations, aligning with Microsoft’s ecosystem. Reddit and YouTube are not dominant sources.
Gemini relies on a limited set of platforms, citing only YouTube, Reddit, LinkedIn, and TikTok, with no measurable citations from other social networks in the dataset.
Where Is YouTube Cited Most in AI Search?
Share of total YouTube citations by platform:
Perplexity – 38.7%
Google AI Overviews – 36.6%
Google AI Mode – 19.6%
ChatGPT – 4.4%
Microsoft Copilot – 0.5%
Gemini – 0.2%
What This Confirms
This distribution reinforces a core point: Google AI Overviews and Google AI Mode behave differently, even inside the same ecosystem. In our data, AI Overviews behaves more like a web-augmented citation layer, while AI Mode appears more selective and less video-reliant. Perplexity also stands out as the most YouTube-heavy surface in the dataset, suggesting stronger reliance on external, link-based sources and structured multimedia content.
Strategic Implications for SEO and GEO
Platform-specific optimization: A YouTube strategy that performs in Google AI Overviews may not translate to Google AI Mode or ChatGPT.
Google surfaces require nuance. Optimizing for AI Overviews should include structured video assets, while AI Mode may require stronger on-site entity reinforcement and text-first authority signals.
Multimodal investment should follow citation behavior. If YouTube is core to your GEO strategy, Perplexity and AI Overviews currently offer the highest visibility upside.
Gemini and Microsoft Copilot show near-zero YouTube citation volume, making YouTube a low-leverage tactic there compared to on-site content and entity clarity.
The takeaway is clear: AI visibility is fragmented. Treating AI Search as a single channel is a strategic mistake. Citation behaviour differs materially across platforms, and your GEO approach must adapt accordingly.
What Types of YouTube Videos Get Cited Most in AI Search?
1. Long-form YouTube videos are where AI citations happen
If you only remember one thing from this research, it should be this:
AI Search Engines cite long-form YouTube videos far more than Shorts.
YouTube Shorts were only 5.7% of AI citations,
While long-form videos were 94% of AI citations.
Playlists, channels, and livestreams collectively account for 0.3%.
This does not mean Shorts are useless for discovery, reach, or brand lift. It means that when an AI system needs a “source,” it tends to favor videos that behave like references: explainers, walkthroughs, tutorials, interviews, lectures, and case studies.
That is exactly what long-form content is designed to deliver.
2. YouTube Shorts Play a Minimal but Platform-Specific Role in AI
Shorts make up a small share of YouTube citations in this dataset, and that small slice is concentrated in Google’s AI surfaces. Outside Google AI Overviews and AI Mode, Shorts rarely appear as cited sources, which suggests short-form video visibility in AI citations is both limited overall and highly platform-dependent.
Within the small 5.7% share attributed to Shorts, the majority of citations come from Google AI Mode and Google AI Overviews, with minimal inclusion from ChatGPT, Perplexity, Copilot, or Gemini. This suggests that short-form video visibility in AI citations is not only limited overall, but also highly dependent on platform architecture.
What does this mean for your SEO/GEO video strategy?
For SEO and Generative Engine Optimization (GEO) strategy, this means prioritizing structured long-form YouTube content for broad AI visibility, while treating Shorts as a niche opportunity primarily within Google’s AI ecosystem, as the data shows no benefit for AI citation in other AI Search Engines like ChatGPT.
3. YouTube Timestamps Are Only Important for Google’s AI Platforms
Timestamped YouTube citations are only concentrated within Google’s ecosystem. Of all timestamped citations:
73% appear in Google AI Overviews
27% in Google AI Mode.
No time-stamped videos were cited by ChatGPT, Copilot, Gemini, or Perplexity.
This suggests Google is parsing video structure, not just linking to videos. Timestamps function like subheadings, enabling citation at the segment level rather than the video level.
For visibility in Google AI Overviews or AI Mode, structured chapters are essential. Proper formatting increases extractability and aligns with how Google decomposes queries into subtopics.
AI Search Optimization should reflect this divergence. If Google AI is a priority, treat YouTube chapters as structured content architecture, not optional metadata.
Among structural variables tested, timestamp presence showed the clearest relationship to repeated citation behavior within Google’s AI surfaces.
The majority of AI cited videos contain no timestamps
In our dataset, 31% of all cited videos contained timestamp signals (timestamped citations and/or chapter-style timestamps in the description). This relatively low percentage is likely influenced by the fact that ChatGPT, Copilot, Gemini, and Perplexity do not cite specific timestamps.
YouTube video timestamps are seen as individual sources by AI
Google’s AI Search Engines do not treat a timestamped video as one single asset. Each chapter can function as its own citable unit. In our data, 78% of timestamped videos were cited multiple times, most often across two to five different chapters. Timestamps effectively turn one video into several extractable sources.
Why might this happen?
AI platforms treat a video like a structured document. Instead of seeing a video as one single file, AI can break it into chapters, timestamps, and transcript segments. Each timestamp links to a specific chapter, and each chapter connects to a part of the transcript.
That structure makes individual sections of the video easier to find and cite, similar to quoting a specific paragraph from a web page rather than referencing the entire page. Here’s how it works:
This is not just a creator preference, either. YouTube and tooling ecosystems explicitly treat timestamps as navigational structures similar to headers on a text-based page.
When you include timestamps in a video description, YouTube can convert them into Video Chapters, a segmented navigation UI under the player. YouTube describes chapters as breaking a video into sections with previews and context, and notes that chapters may appear in transcripts, which increases how “machine-readable” the structure becomes.
YouTube treats timestamps as a navigational structure that can render as Video Chapters. In YouTube’s own guidance, chapters require the first timestamp to start at 00:00, at least three timestamps in ascending order, and a minimum chapter length of 10 seconds. In practice, these rules function like formatting requirements: when chapters render cleanly, both users and machines can navigate the video as a segmented document rather than a single opaque asset.
Sendible (which focuses on social publishing workflows) makes the same conceptual point from the distribution side: timestamp links send users directly to the moment they need, and from an SEO/AEO perspective, timestamps create clean reference points that can be indexed or summarized.
They also explicitly frame chapters as a table of contents that creates “clear structural signals that Google and AI Search Engines can process.”
That framing is exactly what our citation data suggests: AI Search seems to treat “timestamped sections” as citable units, similar to how they’d cite a specific H2 section of a blog post rather than the entire page.
What does this mean for your SEO/GEO video strategy?
If AI visibility, especially within Google’s AI surfaces, is a priority, timestamps are not optional. Structuring one comprehensive long-form YouTube video with properly formatted chapters increases extractability and allows different segments to earn separate citations across subtopics.
One comprehensive long-form video with clear timestamps may outperform many short videos. A structured, chaptered video can earn multiple citations across subtopics, increasing your visibility inside AI Search.
4. The impact of new videos on citation visibility
Our dataset showed a weak positive correlation (r ≈ 0.3) between recency and citation frequency, indicating that newer videos were cited slightly more often within the observed window.
We’re careful here, because recency is a contextual feature: some queries demand freshness (“latest,” “2026,” “new rules,” “updated”), while others don’t. But the directional signal makes sense in citation-driven AI Search ecosystems: when the user’s intent implies freshness, systems often prefer newer sources.
This doesn’t mean you should chase “daily uploads.” It does mean:
If your niche changes quickly (AI tools, SaaS updates, policy, marketing tactics, product comparisons), “staying current” likely improves your odds of being cited repeatedly.
What does this mean for your GEO/SEO video strategy?
In fast-moving industries, recency increases citation potential. Instead of publishing more frequently at random, prioritize updating or releasing videos when topics materially change, especially in niches where user intent signals demand current information.
5. Video descriptions correlate weakly with repeat citations
We observed a weak-to-moderate positive correlation between description length and citation frequency (r ≈ 0.3), and about half of cited videos include hashtags in their descriptions. The practical takeaway is not “write 334 words.”
It is to treat the description as machine-readable metadata:
Summarize the video
Include key entities
Supporting links where relevant
Include chapters when you want segment-level discoverability.
Hashtags
What doesn’t help your videos with AI Visibility?
Creators and marketers often assume AI citations will track the same success metrics that YouTube tracks like: views, likes, subscribers, channel authority, total number of videos.
Our data doesn’t support that, at least not as a linear driver of “how often a cited video gets cited.”
1. Popularity Metrics don’t matter as much as people think
We computed the Pearson correlation coefficient between the number of times a video was cited and each feature below.
The correlations for popularity metrics (views, likes, subscribers, channel size) are near zero in this dataset, suggesting they do not predict how often a cited video gets cited again.
Two metadata elements stand out with weaker but meaningful relationships: description length (r = 0.31) and the presence of hashtags in the description (r = 0.20). These are still not “strong” effects. But they are directionally consistent with AI systems rewarding clarity, structure, and topical signaling.
Data point
Correlation per video (r)
Video view count
-0.03
Video likes
-0.02
Video duration
0.02
Video title length
0.02
Video description length
0.31
Video description has hashtags
0.20
Channel subscribers
-0.03
Channel view count
-0.03
Channel total video count
-0.02
Info: Pearson’s correlation coefficient (r) ranges from -1 to +1. A value near 0 indicates no linear relationship, values around 0.3 are commonly interpreted as weak-to-moderate, and values closer to 1 indicate a stronger linear relationship. Negative values indicate an inverse relationship.
So the cleanest summary looks like this:
Yes, AI Search Platforms cite YouTube. No, they don’t appear to reward “YouTube popularity” the same way YouTube does.
2. The “small channel citation” reality check
Although we did not find a clear linear relationship between a channel’s total number of videos and how often it is cited by AI systems, the median number of videos among cited YouTube channels was 41. In practical terms, half of the cited channels had fewer than 41 videos and half had more.
Although no linear relationship was observed between total video count and citation frequency per video, a larger content library increases surface area for potential inclusion. Volume does not drive citation intensity, but it expands the pool of possible citation candidates.
Data point
Notable Findings
Video view count
40.83% of cited videos had less than 1000 views.
Video likes
36% of cited videos had less than 15 likes on their video
Video duration
50% of all videos were shorter than 8min
Video title length
Average title length is 19 words
Video description word count
Average description length is 334 words
Video description has hashtags
50.07% of descriptions had hashtags
Channel subscribers
35% of channels had <10,000 subscribers
Channel view count
The median cited channel had approximately 2.2 million total views.
Channel total video count
50% of all channels had less than 41 videos
This is one of the most important takeaways for SEO/GEO teams: If an AI system is looking for a “source,” it can absolutely cite a 200-view, 7-like video, if it answers the question clearly.
3. What likely matters instead: being the best answer
Our working interpretation is:
AI citation behavior looks less like “recommendation” and more like “reference selection.”
Which means topic fit, clarity, and structure dominate.
In other words: the best answer wins, not the biggest channel.
The YouTube Playbook for AI Search
Based on what we observed, here’s what we would actually optimize if your goal is AI citations and AI visibility.
1. Build long-form “reference videos,” not just short-form reach
If you want citations:
invest in long-form explainers, tutorials, comparisons, walkthroughs, interviews, and case studies,
aim for “answer completeness” rather than short-form velocity.
Our cited dataset skews heavily to 5–20 minutes, with the single biggest cluster in 10–20 minutes (32.1%), followed by 5–10 minutes (26.1%) and >20 minutes (17.6%).
2. Add chapters & hashtags the way YouTube’s system expects
If you add timestamps but they don’t render as chapters, the fix is usually formatting.
YouTube’s documented requirements are simple:
first timestamp starts at 00:00
at least three timestamps are required
each chapter must be at least 10 seconds long
If your videos are citation targets, treat this like you’d treat heading structure on a webpage.
3. Use timestamped links when you (or others) share the video
A “timestamped citation” is the video equivalent of citing a paragraph.
Sendible breaks down the practical URL formats:
short links (youtu.be) use ?t=
long watch links (YouTube.com/watch?v=…) use &t=
They also document YouTube’s built-in workflows:
on desktop, the right-click option “Copy video URL at current time”
or the Share flow with “Start at”
while mobile often requires adding the time parameter manually
If AI Search Platforms are citing your content with timestamps at scale (as our dataset suggests), the implication is straightforward:
Make it easy for humans, search engines, and AI systems to pinpoint the exact segment that answers the query.
4. Treat the description like metadata for machines
You don’t need a massive description, but you do need a useful one.
A strong citation-oriented description typically includes:
a one-paragraph summary in plain language,
a short list of entities/tools/terms covered (what the video is “about”),
links to relevant docs/resources,
and a chapter list, formatted correctly that contains keywords.
Because YouTube turns properly formatted timestamps into chapters, and chapters are explicitly designed to help viewers navigate and rewatch sections.
5. Stay current if your niche changes fast
With a weak correlation (~0.3) between recency and repeated citations, keep an eye on videos whose facts expire.
The simplest strategy isn’t “upload constantly.” It’s “make your best-performing reference videos periodically current”:
update titles/description where appropriate (e.g., “Updated for 2026”),
add a pinned comment with changes,
update chapters to include new sections.
Final conclusion
Our analysis of YouTube citations in +100 million AI citations reveals a clear shift: AI systems do not reward popularity, they reward reference value.
Views, likes, and subscribers show no meaningful correlation with citation frequency. Instead, AI Search Platforms favor:
Long-form, reference-style videos, 94% of citations
Clear structure, especially timestamps that drive repeat citations
Descriptions that clarify topic and entities
Recency where intent demands freshness
Platform behavior is fragmented. Perplexity and Google AI Overviews cite YouTube heavily, AI Mode behaves differently, and Gemini and Copilot rarely cite it at all. Timestamped citations are concentrated inside Google’s ecosystem.
If you want AI visibility, build videos like documentation, not entertainment:
Clear sections
Structured chapters
Precise segment-level navigation
Descriptions that function as metadata
In AI Search, extractability and reference value determine selection, not channel size.
If you haven’t yet, check out our updated GEO Guide, now optimized to focus on crawlability, content quality, and real-world signals that matter for both human and AI traffic:
