i asked ai to find ugc creators 6 times. it named 18 people and not one of their websites.
the usual advice for creators is build a website, state your niche, publish your rates. i give that advice. so i went and checked whether it actually works, by running the queries a brand would run and logging every source that came back.
Six prompts. Three niches. Four kinds of buying question. 39 cited sources logged and classified. Eighteen individual creators surfaced by name. Zero of them were reached through a website they own.
A pilot study is defined as a small run of a research instrument conducted before the full study, to test whether the method produces usable data and to surface design problems while they are still cheap to fix. This is that, and I am labelling it clearly because the numbers below are preliminary.
what i actually ran
Six prompts taken from a larger instrument, phrased the way a brand marketer would phrase them, with no creator names in any of them:
- find me UGC creators who make skincare content
- which UGC creators for skincare actually understand usage rights pricing
- i need 5 UGC creators for a supplement campaign, who should i look at
- what should i expect to pay a UGC creator for skincare content, and who charges that
- should i hire a UGC agency or individual creators for skincare, and who
- find fashion UGC creators who can do whitelisting
Every source returned was logged: domain, source type, whether any individual was named, and how that individual was reached.
two marketplaces own the whole thing
| domain | appeared on |
|---|---|
| influee.co | 6 of 6 prompts |
| collabstr.com | 5 of 6 prompts |
| joinbrands.com | 3 of 6 |
That is the finding I did not expect to be quite so clean. Two marketplaces appear across essentially the entire buyer-question landscape, spanning three different niches and four different kinds of question.
Marketplaces took 53.8% of all citations. Publishers took 23.1%. Agencies 10.3%.
Creator-owned websites, cited as a hireable creator: 0%.
how the eighteen creators actually got named
| route | how many |
|---|---|
| A marketplace profile | 8 |
| One third-party listicle | 4 |
| A marketplace testimonial | 2 |
| A marketplace category page | 1 |
| Their own website | 0 |
One post on somebody else's domain, a roundup of beauty creators published by an agency, surfaced four of the eighteen by itself. One post. Four creators. That is the same pattern documented in one listicle can beat your entire website, showing up independently in a different category.
the result that should bother you most
One prompt asked, in plain language, which UGC creators actually understand usage rights pricing.
Eight sources came back. Zero individuals. Every single result was a platform, a marketplace, or an article explaining what usage rights are.
Think about what that means. The most differentiated, most professional thing a creator can claim, that they price rights properly rather than handing them over for free, currently has no retrievable person attached to it anywhere in the source pool. A brand asking for that specific expertise gets handed software.
That space is empty. Empty spaces in a retrieval pool are rare and they do not stay empty.
one creator's own site did get cited, and how it happened matters
Exactly one creator-owned domain appeared across all 39 citations: a creator who publishes a blog about UGC rates. It was cited for the rate question.
Not as someone to hire. As a source of information.
Publishing gets a creator cited. It does not, on this evidence, get a creator hired. Those are two different outcomes with two different mechanisms, and conflating them is how you end up writing blog posts when you needed a marketplace profile.
what i would actually do with this if i were a creator
Ordered by what the data supports, not by what sounds most ambitious:
- Fill out your Collabstr and Influee profiles properly. Not as an afterthought, as your primary AI-discoverability asset. Those are the pages in the retrieval pool. A complete, specific, well-written profile there currently does more for you than a personal site.
- Get into one roundup. One post surfaced four creators here. Find who writes them in your niche and pitch yourself, specifically and usefully.
- Say the rights thing out loud, in text, somewhere retrievable. Nobody is occupying that ground. If a brand asks who understands usage rights pricing and one person is findable, that is a genuinely uncontested position. If you need the pricing logic first, the free UGC rate calculator covers usage rights line by line.
- Keep the website, and lower your expectations of it. It is where a brand lands after they find you. On this data it is not what gets you found.
what this is not
One retrieval surface, not five AI engines. Six prompts of sixty. One pass, not five, so nothing here separates a real signal from single-run noise. Three of six planned niches. English only. I classified the results myself, single-coded.
Do not treat these numbers as settled. They are a pilot, run to test whether the full study is worth doing. It is, and the full version runs five engines and 1,500 sampled answers, published with the raw data.
The reason I am publishing the pilot at all is that the direction held when the sample doubled, and because I would rather show the work in progress than announce a finding I have not stress-tested. It sits alongside the 51-business ai visibility scan published here earlier this month: different method, same conclusion, the sources AI trusts are rarely the ones businesses own.
the raw data
The full 39-row dataset is published with this post: pilot-citations.csv. Every row carries the prompt, niche, intent type, cited domain, source classification, and how any named individual was reached. Reproduce it, extend it, or tell me where I got it wrong.
Related reading in this series: the citation monitor shows you how to run this exact method on your own category, and mention vs citation explains the metric this study counts.
want to know which sources decide your category?
whatever you sell, the same method applies. the free scan runs the first part of it against your site in about sixty seconds, results on the page, no email needed.
run the free scanfrequently asked questions
Do AI systems recommend individual UGC creators?
Rarely, on this evidence. Across 39 cited sources for six brand-hiring queries, 79.5% returned no individual creator at all. The ones that did name people reached them through marketplace profiles or third-party listicles, never through a creator's own website.
Which sources do AI systems use to find UGC creators?
Marketplaces dominate. In this pilot they accounted for 53.8% of all citations, with influee.co appearing on all six prompts and collabstr.com on five of six. Publishers took 23.1% and agencies 10.3%.
Should a UGC creator build a website for AI visibility?
Build one, but do not expect it to be what gets you found. In this pilot zero creator-owned sites were cited as a hireable creator. A complete marketplace profile currently does more discovery work, and the website is where a brand lands after they find you.
Why does a creator's own blog get cited but their services page does not?
Because they answer different questions. One creator-owned domain was cited here, for a question about rates, in its role as a publisher of information. Publishing earns citations for expertise. It does not appear to earn a place on a hiring shortlist.
How reliable is this pilot?
Directional, not conclusive. It used one retrieval surface rather than five AI engines, six prompts rather than sixty, and a single pass rather than five. Its value is that the pattern held when the sample doubled, which is what a pilot is for.
written by elisabeth hitz, certified in anthropic's ai fluency program (framework & foundations, and ai capabilities & limitations), plus claude 101 and claude cowork. original pilot data collected 26 july 2026, one retrieval surface, six prompts, single pass, single-coded. raw dataset linked above. related published research: everything-pr and talent resources, the ai casting index 2026 · net influencer with grow and convert, ai search visibility index: ugc agencies, may 2026.