On this page
- The short version
- Who gets the YouTube citations, by industry
- Earlier studies
- How I did it
- 01Businesses' own channels took 68% of the YouTube citations for legal searches and 6% for gaming
- 0259% of the AI Overviews cited a YouTube video, from 39% for beauty to 80% for home improvement
- 03Half of the 2,929 YouTube citations went to 262 of the 1,146 channels
- What I would test
- Limits
- Data
- Cite this study
- About the author
Search Google for "best criminal lawyer": in the first run, the YouTube links in its AI Overview went to three law firms' own channels, Varghese Summersett PLLC, Law Office of John Guidry, P.A. and Tully Rinckey. Search it for "how to game share on ps5": in the first run, five of its AI Overview's YouTube links went to one gaming creator, YourSixStudios, and the sixth to another creator. The same kind of AI Overview links to YouTube both times, but the channels behind the links could hardly differ more. Earlier studies counted how often AI answers cite YouTube; Jellyfish found that independent creators outrank brand-owned content and celebrity influencers in AI assistants' answers, but the article gives no shares by industry and does not cover Google's AI Overviews. So I ran 588 real US searches in 20 industries through Google, three times each, and sorted every YouTube video the AI Overviews cited by who is behind its channel.
The short version
58% of the 2,929 YouTube citations in Google's AI Overviews went to independent creators, 26% to brands' own channels, 9% to media companies and 4% to institutions. Industry by industry the split moves a lot: businesses' own channels took 68% of the YouTube citations for legal searches and 6% for gaming.
A snapshot shows whose videos Google's AI Overviews cite, not what makes them cite one, so the points below are bets to test, not rules. Under each one are the numbers behind it and the finding they come from. The full version, with the strongest objection to each bet, is in What I would test.
- BetIn legal, insurance, real estate, baby and SEO, test your own channel.Evidence in Finding 1:
- Businesses' own channels took 68% of the YouTube citations for legal searches, 51% for insurance, 51% for marketing and SEO, 50% for baby and kids and 50% for real estate and mortgages.
- For legal searches, the median cited channel had 1,860 subscribers.
- BetIn gaming, personal finance, beauty, consumer tech and fashion, test working with the creators AI Overviews already cite.Evidence in Finding 1:
- Creators took 86% of the YouTube citations for gaming, 80% for personal finance, 78% for beauty, 78% for consumer tech and 75% for fashion.
- The table below names the leading channels in each industry.
- BetFor how-to topics, test a video.Evidence in Finding 2:
- 73% of the AI Overviews for how-to searches cited a YouTube video, against 55% for buying searches and 49% for questions.
Who gets the YouTube citations, by industry
For each industry: the share of AI Overviews that cited a YouTube video, then who was behind the cited channels, the median cited channel's size and the channels with the most citations over the three runs. Sorted by the share that went to businesses' own channels.
| Industry | AI Overviews citing YouTube | Brands' own channels | Independent creators | Media | Institutions | Median channel (subscribers) | Leading channels |
|---|---|---|---|---|---|---|---|
| Legal | 54% | 68% | 12% | 9% | 3% | 1,860 | Speaks Law Firm, Iacobelli Law Firm - Personal Injury Law, Ted Spaulding |
| Marketing and SEO | 65% | 51% | 46% | 0% | 0% | 127,000 | Ahrefs, Nathan Gotch, Simplilearn |
| Insurance | 50% | 51% | 18% | 20% | 7% | 12,300 | MoneyGeek, Peter Roberts, The Annuity Expert | Retirement And Insurance |
| Baby and kids | 64% | 50% | 35% | 0% | 10% | 30,600 | Babylist, Graco |
| Real estate and mortgages | 53% | 50% | 35% | 8% | 0% | 28,300 | Rachel Cruze, US Realty Training |
| Tools and home improvement | 80% | 34% | 58% | 4% | 0% | 23,200 | Flooring Insights, Test Mode Tools, Lowe's Home Improvement |
| Home appliances | 66% | 29% | 63% | 5% | 0% | 110,000 | Dyson, Just A Dad Approved, Unsharpen |
| Outdoor and sports | 57% | 29% | 49% | 9% | 10% | 29,800 | REI, Electric Bike Report, Yammie Noob |
| Business software | 70% | 21% | 56% | 11% | 2% | 12,700 | CRM Central, Benjamin Preston, Consumer Desire, George Vlasyev |
| Cars | 67% | 19% | 73% | 4% | 0% | 71,700 | Car Help Corner, Theory of Simple - Travel Prepared For Anything! |
| Fashion and activewear | 49% | 16% | 75% | 8% | 0% | 95,800 | Evumies, Angela Mashelle, Marnie Goldberg |
| Fitness and supplements | 64% | 16% | 70% | 6% | 6% | 347,000 | Jeremy Ethier, Myprotein, TreadmillReviewGuru |
| Pets | 48% | 13% | 65% | 19% | 0% | 84,300 | Canine Revolution Dog Training, Consumer Research Studios |
| Personal finance | 50% | 12% | 80% | 6% | 3% | 106,000 | Steve | Call to Leap, Ali Abdaal, Nischa |
| Travel | 54% | 11% | 67% | 8% | 10% | 32,100 | U.S. Passport Service Guide, Consumer Research Studios |
| Health | 60% | 10% | 45% | 12% | 30% | 325,000 | Tone and Tighten, Sleep Doctor |
| Cybersecurity | 64% | 10% | 53% | 34% | 0% | 63,400 | Safety Detectives, VPNpro, Cybernews |
| Beauty and skincare | 39% | 9% | 78% | 13% | 0% | 334,000 | Blowout Professor, TaraLee |
| Consumer tech | 67% | 7% | 78% | 10% | 0% | 67,100 | History of Simple Things, Ray Tube Fixes |
| Gaming | 62% | 6% | 86% | 8% | 0% | 238,000 | YourSixStudios, Iron Point Tech, GuideRealm |
The rest of each row's YouTube citations went to channels I could not place from what the channel showed: 4% of all YouTube citations.
What earlier studies found
- OtterlyAI found that 94% of YouTube citations across six AI platforms go to long videos, and Perplexity and Google's AI Overviews carry most of YouTube's citations, but it did not split YouTube citations by industry or by who owns the channel, and it ranked no channels (OtterlyAI).
- Jellyfish found that independent creators outrank brand-owned content and celebrity influencers in AI assistants' answers; the article gives no shares by industry and does not cover Google's AI Overviews (Adweek).
What's new here: who is behind every YouTube video Google's AI Overviews cited, labelled brand, independent creator, media or institution, industry by industry and for three kinds of search, with the leading channels in each industry named.
How I did it
- The searches 588 real US searches in 20 industries: 200 buying searches, 193 how-to searches and 195 questions, picked by a fixed rule from keyword data: in each industry, the most-searched searches of each kind, alternating between three starting words, and read one by one to drop off-topic ones. Business software has 22 searches and marketing and SEO 26, because the keyword data ran out of distinct real searches there; every other industry has 30.
- Google Each search three times, US, English, desktop: the first two an hour apart on 12 October and the third about 18 hours later, after the collection stopped overnight.
- A YouTube citation A link to a YouTube video in an AI Overview, counted once per video per AI Overview.
- The channels Each cited video's channel, with its name, description and subscriber count. Claude labelled every channel from those and the titles of its cited videos, without seeing the industry or the search: a brand (a company or business that sells something, including law firms, clinics and agencies), an independent creator, a media company or publisher, or an institution (government, hospitals, universities, nonprofits).
- The check A second Claude labelled a random sample of the channels blind; the labels matched on 97% of them.
Finding 01Businesses' own channels took 68% of the YouTube citations for legal searches and 6% for gaming
- Businesses' own channels took about half or more of the YouTube citations in five industries: 68% for legal, 51% for insurance, 51% for marketing and SEO, 50% for baby and kids and 50% for real estate and mortgages.
- In legal searches the cited channels were small: the median cited channel had 1,860 subscribers, against 59,700 across all YouTube citations.
- Creators took 86% of the YouTube citations for gaming, 80% for personal finance and 78% for beauty.
- Media companies took 34% of the YouTube citations for cybersecurity, where VPN and antivirus review publishers led, and institutions 30% of those for health.
- By kind of search, brands took 28% of the YouTube citations for how-to searches and 19% for buying searches; media took 14% for buying searches.
Finding 0259% of the AI Overviews cited a YouTube video, from 39% for beauty to 80% for home improvement
- By kind of search: 73% of the AI Overviews for how-to searches cited a YouTube video, 55% for buying searches and 49% for questions.
- Google showed an AI Overview on 98% of the 1,764 searches.
- 17% of the YouTube citations went to Shorts.
Finding 03Half of the 2,929 YouTube citations went to 262 of the 1,146 channels
- The median cited channel had 59,700 subscribers; a quarter had 10,200 or fewer.
- The cited videos changed from run to run: 15% of the search and video pairs the AI Overviews cited were cited in all three runs, and 43% in only one.
What I would test
A snapshot shows whose videos Google's AI Overviews cite, not what makes them cite one, so each point below is something to test rather than a rule.
In legal, insurance, real estate, baby and SEO, test whether your own videos get cited. Businesses' own channels took 68% of the YouTube citations for legal searches and 51% for insurance, often from small channels: the median cited legal channel had 1,860 subscribers. The test is a video answering one of your customers' searches, checked in that search's AI Overview a few times over the following weeks.
In creator-led industries, test working with the channels AI Overviews already cite. Creators took 86% of the YouTube citations for gaming and 80% for personal finance, and the table names the leading channels. The test is whether a mention in one of those creators' videos shows up when the AI Overview cites it.
For how-to topics, test whether a video gets cited. 73% of the AI Overviews for how-to searches cited a YouTube video, against 49% for questions. The test is the same searches rerun after the video is up.
What this study does not show
Citations are not clicks. This counts whose YouTube videos Google's AI Overviews cited, not how many people watched them or bought anything after.
The searches are 588 real US searches in 20 industries: 200 buying searches, 193 how-to searches and 195 questions, from a desktop browser, three times each. Business software has 22 searches and marketing and SEO 26, because the keyword data ran out of distinct real searches there; every other industry has 30. Other searches in the same industries may cite other channels.
The labels come from what each channel showed: its name, description, subscribers and the titles of its cited videos. A channel can be hard to place: a lawyer's personal channel counts as a creator unless it presents a firm, and 4% of the YouTube citations went to channels that fitted no label. Claude labelled the channels and a second Claude checked a sample blind; neither is a person checking who owns the channel.
The runs were not evenly spaced: the first two an hour apart on 12 October and the third about 18 hours later, after the collection stopped overnight. The cited videos changed from run to run, so a single check of one search can show different channels.
Data
1,764 Google searches (588 searches, three times each), 1,727 of them with an AI Overview. Those AI Overviews held 2,929 YouTube citations of 1,523 videos from 1,146 channels. Every percentage has a numerator and a denominator in the study's metrics file, every channel label its reason, and every Google result is kept as the raw response.
Cite this study
Ivan Smiljkovic (2026). 58% of the YouTube Citations in Google's AI Overviews Go to Independent Creators; in Legal Searches, 68% Go to Businesses' Own Channels (Study of 2,929 YouTube Citations). Ivan Builds. https://ivanbuilds.com/youtube-ai-overviews-by-industry.html
Every section has its own link: hover a heading and use the # to copy it.
About the author
I study how Google and AI assistants choose what to cite, using public data I collect myself, and I publish the method and limits with every study.