YouTube Is the Top-Cited Domain in Google's AI Answers. In ChatGPT It's 1%.
Six published figures for YouTube's presence in AI answers run from 0.99% to 23.3%, and every one of them is defensible. The spread is engines and denominators — and it decides whether video belongs in your AEO plan at all.
Updated · Published
Frequently asked questions
Is YouTube the most-cited domain in AI answers?
On Google's AI surfaces, yes, by every published measure. Surfer's analysis of 46 million AI Overview citations put YouTube first at about 23.3%, and Ahrefs' July 2026 tracking across more than 3 million US queries put it first at 21.1% of top-source citations. Across chat assistants the picture is weaker: Meltwater found YouTube first at 1.52% of 9.5 million citations, and in ChatGPT specifically YouTube runs at about 1% of citations.
Why do YouTube citation figures range from 1% to 23%?
Because the studies measure different engines with different denominators. A share of all citations across six platforms (Meltwater's 1.52%) is not a share of the citations earned by the top 50 domains in AI Overviews (Ahrefs' 21.1%), and neither is the share of answers containing at least one YouTube link (Bluefish's 16%). The figures are consistent once the denominators are lined up.
Does ChatGPT cite YouTube?
Rarely, relative to Google. Profound data reported by Marketing Dive put YouTube at 0.99% of ChatGPT citations against Reddit's 2.4%. Bluefish data reported by Adweek found YouTube in 16% of LLM answers over six months, but that is a response rate across assistants, not a ChatGPT citation share. On the engine with the most users, video is a minor source.
Did YouTube really overtake Reddit as an AI citation source?
On some surfaces. Bluefish data reported by Adweek in January 2026 put YouTube in 16% of LLM answers against Reddit's 10%, and Ahrefs' July 2026 AI Overviews data has YouTube at 21.1% against Reddit's 18.5%. In ChatGPT the ordering still runs the other way, with Reddit's citation share more than double YouTube's. The overtake is real but engine-specific.
Does a video need to rank in Google to be cited in AI Overviews?
No. Ahrefs' March 2026 study found that among cited pages that did not rank in Google's top 100 for the query, 18.2% were YouTube URLs — the single largest group. Query fan-out retrieves videos against sub-queries the user never typed, so a video can earn citations while ranking for nothing in the text SERP.
Why does Google's AI cite YouTube so heavily?
Nobody has published a confirmed mechanism, and at least three explanations fit. Google owns YouTube, and Surfer's data also has google.com itself at 16.4% of AI Overview citations, which reads as self-preferencing. Transcripts and chapters make videos unusually extractable. And fan-out retrieval surfaces videos for sub-queries where text pages are thin. The three are not exclusive, and none of them transfers to ChatGPT.
How do you make a video citable by an AI engine?
Write for the transcript, because engines read text rather than watching video. Say the answer out loud in extractable sentences near the start, upload a clean transcript instead of relying on auto-captions, add chapters so the engine has section boundaries, and put specifics — numbers, versions, prices — in the description. A video whose substance lives on screen rather than in the audio is invisible at retrieval time.
Does YouTube's AI citation share vary by industry?
Sharply. Surfer's per-industry figures for AI Overviews run from roughly 23% of citations in finance to 32% in e-commerce and 39% in SEO topics, with gaming reported at over 90%. The spread is wide enough that no cross-industry average should move a video budget — your category's number is the only one that matters, and it is readable from your own tracked prompts.
Should we shift AEO budget into YouTube?
Only if your tracked prompts show Google's AI surfaces answering your buyers' questions. Where AI Overviews and AI Mode carry your category, video is the most underweighted asset in most content plans. Where your buyers ask in ChatGPT, the same investment buys roughly 1% of the citation pool. Check the engine mix in your own data first.
There are at least six published numbers for how often AI answers cite YouTube, and they span 0.99% to 23.3%. Surfer counted 46 million AI Overview citations and put YouTube first at about 23.3%. Meltwater counted 9.5 million citations across six platforms and put YouTube first at 1.52%. Profound data reported by Marketing Dive put it at 0.99% of ChatGPT citations — not first at all. Every figure is sound. The spread is engines and denominators, and sorting it out is the difference between a video strategy and a video mistake.
Key facts
Google AI Overviews, share of citations
23.3% — rank 1, ahead of Wikipedia at 18.4% (Surfer, 46M citations, Mar–Aug 2025)
Google AI Overviews, share of top-50 sources
21.1% — rank 1, ahead of Reddit at 18.5% (Ahrefs, Jul 2026)
Google AI Overviews, share of all citations
5.6% (Ahrefs, Mar 2026, 863k SERPs)
ChatGPT, share of citations
0.99% — behind Reddit at 2.4% (Profound)
Six platforms combined
1.52% — still rank 1 overall (Meltwater, 9.5M citations)
AI Overviews vs organic, same queries
YouTube 4.3× more often; Reddit 3.9× the other way (LQ Digital)
Cited pages ranking nowhere in Google's top 100
18.2% were YouTube URLs — the largest single group
On Google's AI surfaces, YouTube is the top-cited domain in every dataset that isolates them: 23.3% of citations in Surfer's March–August 2025 corpus, 21.1% of top-source citations in Ahrefs' July 2026 tracking.
Across chat assistants it is a top source in a field with no dominant sources: Meltwater's 1.52% ranks first among all domains while remaining a rounding error of total citations.
ChatGPT is the exception that decides budgets: 0.99% of citations, still well behind Reddit's 2.4%.
The widely repeated "YouTube overtook Reddit" finding — Bluefish's 16% of LLM answers versus 10% — is real, engine-dependent, and measured on a different denominator than any citation-share figure.
Videos get cited without ranking: among AI Overview citations that ranked nowhere in Google's top 100, 18.2% were YouTube URLs, the largest single group.
No published study splits brand channels from creator channels. The data says video gets cited; it does not say your videos get cited.
Three sentences carry the whole comparison. On Google's AI Overviews, YouTube is the most-cited domain, at 21.1–23.3% of citations depending on the denominator. On ChatGPT, YouTube is 0.99% of citations, behind Reddit's 2.4%. Across six platforms pooled, YouTube is 1.52% of citations and still ranks first, because no domain on the open-web assistants holds a meaningful share.
Three things generate the spread. The surface: Google's AI answers cite YouTube at a different order of magnitude than the assistants do. The denominator: Ahrefs' 21.1% is YouTube's slice of the citations earned by the top 50 domains only — its own methodology note says mention share is "a domain's citations as a percentage of the summed citations of the top 50 most-cited sources" — while its March study's 5.6% is a share of all citations, long tail included. Same company, same surface, both correct. The metric: Bluefish's 16% counts answers containing at least one YouTube link, which is the same response-versus-citation split that produced LinkedIn's 11%-versus-0.53% pair.
The number relevant to a visibility decision is the one matching your engine and your question. "How much of the AI Overviews citation pool is video?" is the 5.6%. "Who wins the head-to-head among top sources there?" is the 21.1%. "How often does an assistant answer contain a video?" is the 16%.
Whatever else the datasets disagree on, they agree here. Surfer's corpus — 36 million AI Overviews between March and August 2025 — has YouTube first at 23.3%, ahead of Wikipedia at 18.4%. Ahrefs' July 2026 tracking has it first at 21.1%, ahead of Reddit at 18.5%. And LQ Digital's comparison of 8,000-plus citations against organic results for the same queries found YouTube appearing 4.3x more often in AI Overviews than in organic — the mirror image of Reddit, which runs 3.9x the other way, a pattern covered in our review of Reddit's AI citations.
Three explanations fit, and none excludes the others. The first is ownership. Google cites its own video platform at rates no other engine matches, and Surfer's same dataset has google.com itself at 16.4% of citations — the third most-cited domain in Google's own answers is Google. Nobody outside Google can measure intent here, but the benchmark consequence is measurable either way: YouTube's dominance of AI Overviews tells you nothing about engines Google does not run.
The second is fan-out. Ahrefs' March 2026 data found that among cited pages ranking nowhere in Google's top 100 for the query, 18.2% were YouTube URLs. A system that retrieves against generated sub-queries rather than the typed keyword will surface a video that text rankings buried, because the video answers a sub-question no ranked page covers.
The third is extraction: an engine reads a video's transcript and surrounding text rather than watching it, which makes a well-transcribed video one of the most machine-readable sources on the web. That mechanism decides what a video team should actually do, so it gets its own section below.
The vertical spread is wide enough to check before acting on any average:
Industry
YouTube's share of AI Overview citations (Surfer)
Gaming
>90%
SEO topics
39%
E-commerce
32%
Finance
~23%
Your category's number is knowable from your own tracked prompts and is the only one that should move budget.
Meltwater's 9.5 million citations across six platforms rank YouTube first among all domains — at 1.52% of citations. That is what first place looks like on the open-web assistants: no domain holds a meaningful share, and the leader clears the field at one and a half percent.
The Bluefish finding sits on top of this. YouTube in 16% of LLM answers against Reddit's 10%, over six months, reversed the ordering that the entire Reddit-for-AEO thesis was built on. Adweek's report attributes the rise to machine-readable transcripts, though it publishes no engine-level breakdown or methodology, so the number is best read as a direction rather than a benchmark.
ChatGPT is where the direction fails. Profound's figures give Reddit 2.4% of ChatGPT citations and YouTube 0.99% — the one major engine where the overtake has not happened, and the engine with the largest user base. A team that reads "YouTube overtook Reddit" and shifts effort accordingly, while its buyers ask in ChatGPT, has optimized for the wrong machine.
Nobody knows, and that is the most consequential gap in the published data. Every study above counts youtube.com as a single domain, so a citation to a brand's own product walkthrough and a citation to an independent creator's teardown are the same data point.
The distinction decides what you actually do with the finding:
If the cited videos are mostly…
The work is…
The budget line is…
Brand-owned channels
Publishing — your own explainers, demos, and comparisons
Video production
Independent creator channels
Briefing — getting accurate information to the people engines already cite
Creator relations and PR
This is the same split LinkedIn's citations show between personal profiles and company pages, and it resolves the same way: no published dataset answers it, but your own cited-URL data does. Pull the full YouTube URLs cited for your tracked prompts and read the channel names. It is a twenty-minute check that no vendor study can do for you, because the answer is category-specific.
An engine does not watch a video. It reads the transcript, the chapters, the title, and the description — which means the citable unit is the spoken words, and a video whose substance lives on screen rather than in the audio is invisible at retrieval time. Adweek's report puts the mechanism in one line: transcripts, explainers, and the text around videos "have allowed the video platform to flourish as a source that machines can easily read."
That yields four concrete production rules:
Say the answer out loud, near the start, in extractable sentences. If the answer only exists as text on a slide, it does not exist.
Upload a clean transcript. Auto-captions are the worst text you can hand a retrieval system, and they are what you get by default.
Add chapters. They give the engine section boundaries, which is the video equivalent of the H2s this page is using.
Put specifics in the description. Numbers, model names, versions, prices — the things a fan-out query is looking for.
One counterintuitive consequence: do not grade citation-earning videos on views. Retrieval and recommendation are different systems, and a video can feed AI answers for months while looking dead in YouTube Studio.
Surfer, Ahrefs, and Meltwater sell tooling adjacent to the conclusion; Bluefish and Profound sell AI visibility services and published through trade press without full methodology. Surfer's report states no limitations at all. Ahrefs' most-cited list is regenerated monthly by its own automated agent, which keeps it current and makes any quoted figure a snapshot. The direction of the findings survives this; the second decimal place does not.
A YouTube citation under an answer does not mean the answer named your brand — the ghost citation gap applies to video sources the same as any other. And a citation to a comparison video naming four competitors is not obviously a win.
Bluefish's measurement is a six-month window ending January 2026; Surfer's corpus ends August 2025; Ahrefs updates monthly. Given how fast cited sources turn over, treat every figure here as a reading, not a constant.
Pull the cited YouTube URLs from your own tracked prompts, per engine. The published spread — 21% on Google's surfaces, 1% on ChatGPT — is wide enough that no cross-engine average supports a decision.
Look at the channel names in those URLs. Brand channels mean you should publish; creator channels mean you should brief. No study makes this call for you.
If AI Overviews carry your category, treat video as underweighted. YouTube appears 4.3x more often there than in organic results for the same queries; most content plans still allocate as if organic were the whole game.
Write for the transcript, following the production rules above — the engine reads text, not video.
Keep Reddit and everything else in the mix on ChatGPT. The overtake headline is engine-specific, and on the highest-volume assistant the old ordering still holds.
Elmo is an open-source, self-hosted AI visibility platform that runs your prompt sets across ChatGPT, Claude, Gemini, Perplexity, and Google's AI surfaces, recording every citation, mention, and competitor named alongside you. Because it stores the full cited URL rather than just the domain, counting YouTube citations per engine — and splitting your channel's videos from creators' — is a query over your own results rather than a manual audit.