Updated October 5, 2026

YouTube appears in roughly one-third of cited Google AI Overviews answers in LLM Pulse’s current data. In ChatGPT, the picture looks completely different.

That difference matters more than any ranking of the “most cited websites in AI.”

When we first researched this question in April 2026, the available evidence pointed strongly toward YouTube as an important source for AI citations. Six months later, there is considerably more data to work with, and it reveals something the earlier research could not show clearly: YouTube’s importance varies dramatically by AI system.

Some of our original conclusions still hold. Others were too broad, and newer evidence lets us correct them. The better question now is not whether YouTube matters to AI visibility. It is: which AI system are you trying to appear in?

The answer changes what you should optimize.

Key takeaways

Is YouTube the most cited source in AI search right now?

In Google AI Overviews, yes, by a wide margin. LLM Pulse runs thousands of prompts a week through ChatGPT, Perplexity, Gemini, Google AI Mode and Google AI Overviews. In its snapshot for the seven days ending October 4, 2026⁠, youtube.com was cited in 9.66% of cited answers, ahead of google.com (6.39%) and reddit.com (6.26%).

Filter to Google AI Overviews⁠ and YouTube appeared in 33.48% of cited answers. Among AI Overview answers containing citations in LLM Pulse’s sample, roughly one in three cited YouTube.

Google AI Mode is closer to a tie. Ahrefs’ September 2026 US data⁠ puts Reddit at 17.9% of AI Mode’s top-source citations and YouTube at 17.8%.

Filter to ChatGPT⁠ and YouTube is not in the top 10 at all.

Why do the “most cited website” rankings disagree?

Because they measure different things. Every citation study picks a set of prompts, a set of engines, a market and a way of counting, and each choice moves the result.

Three datasets published or refreshed in the last five weeks show it:

LLM Pulse, October 4, 2026

Across five AI engines, YouTube ranked first. It appeared in 9.66% of cited answers. LLM Pulse measures the share of cited answers that reference a domain at least once over a seven-day window.

Ahrefs Brand Radar, September 2026

In its US ChatGPT data, Reddit ranked first at 16.8% of citations. YouTube ranked eighth at 2.5%.

Agenzy, September 8, 2026

Across 2,283,923 citations from buyer-focused questions, mostly involving Lithuanian businesses, Reddit ranked first at 2.1% and YouTube second at 1.6%.

LLM Pulse, October 4, 2026

Across five AI engines, YouTube ranked first. It appeared in 9.66% of cited answers. LLM Pulse measures the share of cited answers that reference a domain at least once over a seven-day window.

Ahrefs Brand Radar, September 2026

In its US ChatGPT data, Reddit ranked first at 16.8% of citations. YouTube ranked eighth at 2.5%.

Agenzy, September 8, 2026

Across 2,283,923 citations from buyer-focused questions, mostly involving Lithuanian businesses, Reddit ranked first at 2.1% and YouTube second at 1.6%.

None of these is wrong. Run prompts through Google AI Overviews and YouTube is everywhere. Run US queries through ChatGPT and Reddit and Wikipedia dominate.

Agenzy’s study⁠ adds a second correction. No single domain held more than 2.1% of its citations, and it took 1,057 different domains to account for half of them.

The famous top-10 lists describe a thin layer on top of a very long tail. Our read: that tail is where most small businesses live, and in it, being the clearest answer to a specific question counts for more than being a famous domain.

Which AI platforms actually cite YouTube?

Google’s, mostly, plus Perplexity.

OtterlyAI’s YouTube Citation Study 2026⁠, published March 2, 2026, tracked more than 100 million AI citations over 30 days and split YouTube’s citations by where they appeared:

Platform Share of all YouTube citations

Perplexity 38.7%

Google AI Overviews 36.6%

Google AI Mode 19.6%

ChatGPT 4.4%

Microsoft Copilot 0.5%

Gemini 0.2%

Agenzy’s September data points the same way from the other side. YouTube took 4.7% of Google AI Overviews citations in its dataset and 0.03% of ChatGPT’s.

One caution on Perplexity. Otterly’s number is Perplexity’s share of all YouTube citations, while Agenzy measured YouTube’s share of Perplexity’s citations (0.6%) on a different prompt set.

Different denominators, different prompts: treat Perplexity as strong for YouTube on some query types, not all.

What do AI-cited YouTube videos have in common?

Not audience size.

In Otterly’s dataset, view count and channel subscribers both correlated with citation frequency at r = -0.03, which is effectively zero. 40.83% of cited videos had fewer than 1,000 views.

What showed positive correlations with repeat citation frequency: description length (r = 0.31), hashtags in the description (r = 0.20) and recency (r ≈ 0.3). Otterly itself calls these weak to moderate. The average cited video carried a 334-word description.

Format mattered more than any single metric. Long-form videos took 94% of YouTube citations and Shorts 5.7%, with most of the Shorts citations coming from Google’s surfaces.

One limit applies to all of it. Otterly studied only videos that were already being cited, so these findings describe what gets cited again, not what gets a video cited the first time.

A correlation tells you where to look, not what to do.

How should you structure a video so AI can cite it?

Build it like a reference document: one clear topic, a written description that states what it covers, and chapters that split it into answerable sections.

Chapters carry the clearest evidence. Every timestamped YouTube citation Otterly observed appeared in Google AI Overviews (73%) or AI Mode (27%), and 78% of timestamped videos were cited more than once, often across two to five chapters.

Only 31% of cited videos had timestamp signals at all, so even among videos already appearing in AI citations, structured timestamps were far from universal.

For chapters to render, YouTube needs the first timestamp at 0:00, at least three timestamps, and chapters of at least 10 seconds.

Label each one with the question it answers, not its position in the video:

0:00 What this video covers

1:30 Why auto-generated captions misspell technical terms

4:15 How to write a YouTube description that states what the video answers

8:00 Adding chapters to videos you already published

11:45 Checking whether Google AI Overviews cites your video

“Part 2” tells a retrieval system nothing. “How to add chapters to videos you already published” tells it exactly which question that segment answers.

The description is where you state what the video is about before anything else gets read. Open with the question the video answers, summarize each section, name the tools, standards and organizations involved, and include the chapter list.

Transcripts are the part nobody has measured directly. No study we found isolates transcript accuracy as a citation factor, so this is inference: a system citing a specific segment needs text it can attach to that segment, and captions that mangle your key terms make that harder.

Uploading a corrected transcript costs little and removes the risk.

What if ChatGPT is the AI you care about?

Then YouTube is not your main surface.

Ahrefs’ September 2026 data⁠ puts YouTube at 2.5% of ChatGPT’s top-source citations, behind Reddit, Wikipedia and a run of publishers and reference sites.

It did climb seven places on the prior month, but one month is a data point, not a trend.

ChatGPT’s citations lean on text: forums, reference pages, publishers, and pages that answer the question directly.

That is why every video worth making is worth a matching page on your own site, written from the same transcript and structured around the same questions.

Agenzy found brands’ own websites earned 3.7% of citations against 16.9% for their named competitors. Its reading is that a brand’s site gets cited when it carries a page answering the buyer’s question, which is a structural problem you can fix.

What changed since the April version?

We researched this question in April 2026 using the evidence available then. In October we went back, checked every original claim against its source, kept what survived and corrected what did not.

What held up: long-form video dominates YouTube’s AI citations, audience size shows almost no relationship with citation frequency, and Google’s AI surfaces can cite specific timestamped sections of YouTube videos.

What we corrected: we removed figures we could not trace to a published source, including a claimed 16% share of all LLM answers and a citations-per-100,000-views comparison. We corrected the description-length guidance to Otterly’s reported 334-word average and removed the claim that Gemini is a major YouTube citer, since Gemini accounted for 0.2% of YouTube citations in Otterly’s window. We also scaled back our transcript claims, which were stated more strongly than the evidence supports.

What is new: per-engine data from LLM Pulse, Ahrefs and Agenzy, which shows that YouTube’s importance depends on which AI system you are trying to appear in.

FAQ

What website do AI search engines cite the most in 2026?

It depends on the engine. In LLM Pulse’s October 4, 2026 snapshot across five engines, YouTube appeared in the most cited answers (9.66%). In ChatGPT specifically, Ahrefs’ September 2026 US data ranks Reddit first at 16.8% of citations, followed by Wikipedia at 7.0%.

Does ChatGPT cite YouTube videos?

Occasionally, and far less than Google AI Overviews does. Ahrefs’ September 2026 data puts YouTube at 2.5% of ChatGPT’s top-source citations, and ChatGPT accounted for 4.4% of all YouTube citations in OtterlyAI’s 2026 study. For ChatGPT visibility, your website usually matters more than your channel.

Does Google AI Overviews cite YouTube?

Yes, more than any other domain in current tracking. LLM Pulse found YouTube cited in 33.48% of Google AI Overviews answers that contained citations in the week ending October 4, 2026. Agenzy’s study measured YouTube at 4.7% of all AI Overviews citations on a different prompt set.

Does my subscriber count affect whether AI cites my YouTube videos?

Not in the available data. OtterlyAI found a correlation of r = -0.03 between subscriber count and citation frequency among already-cited videos, and 40.83% of cited videos had fewer than 1,000 views.

Do YouTube chapters help AI citation?

In Google’s AI surfaces, the evidence points that way. All timestamped YouTube citations in OtterlyAI’s study appeared in Google AI Overviews or AI Mode, and 78% of timestamped videos were cited more than once. No other platform cited timestamps during the study window.

Do YouTube Shorts get cited by AI?

Rarely. Shorts took 5.7% of YouTube citations in OtterlyAI’s 2026 study against 94% for long-form, and most Shorts citations came from Google’s AI surfaces.

Should I make YouTube videos or blog posts for AI visibility?

Both, aimed at different engines. YouTube is a strong citation surface for Google AI Overviews and AI Mode, while ChatGPT leans on text sources, so a video paired with a page built from its transcript covers both.

⸻

Every month someone publishes a new list of the websites AI cites most. The useful question was never which site is on top. It was which answer you want to be in, and whether you gave that engine something it can quote.

⸻

Originally published by AI Visibility Studio⁠. AI Visibility Studio audits what AI systems can find, read and cite across your website and content.