YouTube is the most-cited source in AI answers. Which engines does that actually reach?

KEY TAKEAWAYS

1. YouTube's lead in AI citations comes mostly from Google AI Mode, AI Overviews and Gemini. ChatGPT barely links video.
2. Two routes exist. Your own video gets cited by Google surfaces. Your brand named in other people's videos tracks with AI mentions on every engine.
3. Relevance between the question and the title and description predicts citation. Views and subscribers do not.
4. Weight each engine by how your buyers use it before deciding how much video effort AI visibility justifies.

In August 2026, YouTube was the most-cited source across eight AI products, ahead of Reddit, according to Meltwater data reported by Digiday on 23 September. That headline has started a rush to build channels for AI visibility. Read engine by engine, the data says something narrower. YouTube citations come mostly from Google. Whether a channel is worth it depends on which engines your buyers use.

Where do YouTube's AI citations actually come from?

Meltwater's August report tracked about 7.3 million citations from unbranded consumer prompts across eight engines. YouTube was first overall, with 212,500 citations, 7.34% fewer than in July. The per-engine counts show where the lead comes from:

Scrunch measured YouTube citation rates by engine from May to July 2026: about 30% in AI Mode, 27% in AI Overviews, 15% in Perplexity, 4% in Gemini, 0.5% in Copilot and 0.3% in ChatGPT. Scrunch does not define the denominator, so treat these as a ranking, not exact shares. The order matches other trackers. Google's surfaces cite video often. ChatGPT almost never does.

Why do the YouTube share figures disagree so much?

You will see YouTube described as 1.68% of AI citations, 5.6% of AI Overview citations, 31.8% of social citations and first overall. All can be true, because each divides by something different.

• All citations. Promptwatch put YouTube at 1.68% of all citations in February 2026. Ahrefs found YouTube URLs were 5.6% of about 4 million AI Overview citations.
• Social citations only. OtterlyAI's 100 million citation study found YouTube took 31.8% of social and video citations, which were themselves 5.54% of the total.
• Rank by count. Meltwater ranks domains by raw citations. A domain can be first while holding a small share, because citations spread across thousands of sites.

The window matters as well. OtterlyAI's spring data showed Gemini with 0.2% of all YouTube citations. Meltwater's August data showed Gemini as YouTube's largest source. Digiday reported Weglot's finding that Claude and Gemini often embed videos as recommendations. Some trackers count those, some do not. A recommended video is not proof the video shaped the answer.

What changed for ChatGPT and Perplexity this summer?

Both engines moved in 2026, and neither in YouTube's favour.

Goodie tracked 1,856,259 social citations from January to August 2026. It saw ChatGPT drop YouTube entirely on 30 March and bring it back at the end of July. After a further shift in mid-August, YouTube was 4% of ChatGPT's social citations, against 77% for Reddit. Obsurfable's read of Ahrefs data had YouTube falling 11 places in ChatGPT, to 1.8% of mentions among top sources.

Perplexity looked like YouTube's strongest non-Google engine. Ahrefs' July snapshot, via Obsurfable, had YouTube first there at 31.2%. Then Scrunch reported that around 12 August, "Perplexity stopped citing YouTube almost entirely." Meltwater's August data is less extreme but points the same way: fourth place. Perplexity's release notes from July to October mention no change to YouTube sources, so there is no stated cause. The safe reading is that Perplexity's video citations are unstable. Its weight is also shrinking. Search Engine Journal reported its share of AI referrals in StatCounter data fell from 7.91% in June to 4.31% in August.

Is there a route to ChatGPT through YouTube at all?

Possibly, but not through your own video link. Ahrefs studied about 75,000 brands with Brand Radar. The strongest correlate of being mentioned in AI answers was YouTube mentions: your brand named in any video's title, transcript or description. The Spearman correlation was about 0.737 overall, and it held at a similar level for ChatGPT, AI Mode and AI Overviews. View-weighted mentions correlated slightly less, at 0.717.

That points to two separate routes:

1. Your own video as a cited source. This mostly reaches Google AI Mode, AI Overviews and Gemini.
2. Your brand named inside other people's videos. This tracks with mentions on every engine, ChatGPT included.

Ahrefs is clear that this is correlation. Brands that creators talk about are usually brands that the web talks about. Google's own guide to generative AI search says these features can draw on what people say about products across blogs, videos and forums, and warns against seeking inauthentic mentions. In digital authority terms, route one sits with search presence and AI visibility. Route two behaves like press coverage. It is earned mention in sources the engines read.

Scrunch's sponsored-video data adds a qualified point. When an answer cited a sponsored video, the sponsor was named in about 40% of answers, against 32.3% when it did not. That is a 1.24x lift, observational, and only present while the video is being cited.

What makes a video citable?

Not popularity. The studies agree on this more than on anything else.

• OtterlyAI found correlations with views and subscribers of -0.03. 40.83% of cited videos had under 1,000 views.
• Gumshoe compared 5,000 videos cited in AI Overviews with similar uncited ones. A cited video beat its rival on views only about 52% of the time. Similarity between the prompt and the title and description was the strongest single signal.
• Description length correlated at 0.31 in OtterlyAI's data. Gumshoe also found longer descriptions helped, and warned that padding them would not necessarily earn citations.
• Long-form took 94% of YouTube citations in OtterlyAI's data, and Shorts 5.7%.
• Timestamped citations appeared only in Google's surfaces: 73% in AI Overviews and 27% in AI Mode.

Chapters are cheap to add. Google says it prioritises key moments set by timestamps in a YouTube description. YouTube requires the first timestamp at 00:00, at least three timestamps, and chapters of at least 10 seconds.

Title alignment matters when a video and a page cover the same story. In OtterlyAI's one-week test of eight video and blog pairs, the written version took 79.5% of format citations. Where the titles nearly matched, the video took 53%. Where they did not, video got under 3%. The sample is small, so read it as a direction.

Do video citations fade or last?

Both, depending on what you measure. Scrunch followed sponsored videos after their first citation. The citation rate fell from 18.4% to 10.5% within a month, roughly 43%. Promptwatch photographed what was cited in February 2026 and found 54% of cited videos were more than two years old, and about 4% under a month.

These do not conflict. Scrunch measures the flow of newly cited videos. Promptwatch measures the stock. Most new videos get a short run, and a few become the standing answer to a question. That argues for videos that answer durable questions, and for checking at four to six weeks rather than at launch.

One dissent is worth noting. 5W ran 15 buyer-intent queries in five verticals and found no YouTube URLs among 109 citations. The engines are unnamed and the sample is tiny. It still matches a pattern in the other data: commercial comparison prompts often cite vendors and trade press, not video.

A worked example: how much of your answer surface can video reach?

The numbers below are hypothetical. They show the method, not a benchmark.

A B2B software brand runs 30 buyer prompts in each engine and counts the answers that cite any YouTube URL. It also estimates how its buyers split their AI use, from sales calls and a short customer survey.

About 13% of this brand's weighted answer surface pulls a YouTube link today. A consumer brand whose buyers lean on Google, with weights of 10%, 40%, 20%, 20% and 10%, gets about 24% from the same rates. The first brand should retrofit what it has and work on creator mentions. The second has a case for a real channel plan. Prompt counts this small carry wide error, so size your sample before you trust a single run.

Which route fits your situation?

A seven-step check you can run this month

1. List 20 to 40 buyer questions. Use the words customers use, not your keywords.
2. Run each question in every engine you care about. Record every YouTube URL cited and the channel that owns it.
3. Mark citations versus recommendations. A video linked as a source is different from a video embedded as a suggestion.
4. Weight engines by your buyers. Use sales calls, surveys and referral data, and treat the weights as estimates.
5. Score both routes. Count your own videos cited, and count videos by others that name you or a competitor.
6. Fix before you film. Add chapters, rewrite thin descriptions and align titles with the matching page. Verify your channel in Search Console as a platform property, so you can compare watch and Shorts URLs and see which queries reach it.
7. Re-check at four to six weeks. Track which videos keep their citations, and drop formats that fade.

If you want this run across ChatGPT, Perplexity and Google's AI surfaces as part of a digital authority score, join the waitlist.

Questions

Does ChatGPT cite YouTube videos?

Rarely. Goodie found YouTube was about 4% of ChatGPT's social citations after a mid-August 2026 shift, and Scrunch measured a YouTube citation rate of about 0.3% for ChatGPT. Goodie saw ChatGPT drop YouTube entirely from 30 March to late July 2026. Plan your own videos for Google's AI surfaces, not for ChatGPT links.

Do views or subscribers decide which videos AI engines cite?

Not much. OtterlyAI found near-zero correlation with views and subscribers, and Gumshoe found a cited video beats an uncited alternative on views only about 52% of the time. Relevance between the question and the video's title and description was the strongest signal.

Should we make Shorts or long videos for AI visibility?

Long videos. In OtterlyAI's data 94% of YouTube citations went to long-form videos and 5.7% to Shorts. Gumshoe found cited videos ran a median 9.4 minutes against 6.2 for uncited ones, a small effect. Length follows from covering one question properly, not the reverse.

How long does a YouTube citation last?

Often not long. Scrunch saw the citation rate of newly cited sponsored videos fall from 18.4% to 10.5% within a month. Yet Promptwatch found 54% of cited videos were over two years old. Most new citations fade, and a minority of videos become lasting answers.

Sources

1. Meltwater, AI Search Visibility Report August 2026 meltwater.com
2. Digiday, In Graphic Detail. LLMs keep citing YouTube in search results digiday.com
3. Scrunch, YouTube influencers have an AI audience. What millions of citation events reveal about influencer sponsorships scrunch.com
4. Goodie, Social media AI citations. 16 sourcing shifts higoodie.com
5. OtterlyAI, YouTube AI Citation Study 2026 otterly.ai
6. OtterlyAI, Video to Blog. Which format wins in AI search? otterly.ai
7. Gumshoe, What gets a YouTube video cited in AI Overviews gumshoe.ai
8. Ahrefs, Top brand visibility factors in ChatGPT, AI Mode and AI Overviews (75k brands studied) ahrefs.com
9. Ahrefs, Update. 38% of AI Overview citations pull from the top 10 ahrefs.com
10. Promptwatch, What kind of YouTube videos get cited in AI search? February 2026 data promptwatch.com
11. Obsurfable, Top domains cited by LLMs. August 2026 obsurfable.com
12. 5W, Creator visibility online in AI answer engines 5wpr.com
13. Search Engine Journal, ChatGPT, Gemini and Claude lead AI visibility. Is it time to stop tracking Perplexity? searchenginejournal.com
14. Releasebot, Perplexity release notes releasebot.io
15. Google Search Central, Optimizing for generative AI search developers.google.com
16. Google Search Central, Video SEO best practices developers.google.com
17. Google Search Central, Analyze social and video platform content developers.google.com
18. YouTube Help, Video chapters support.google.com

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →