Where do you find UGC creators for an AI app? First decide which ad you want made.

I am Samet Durgun, a fractional Head of UA. I run paid UA for subscription apps and mobile games and write up what I find in the accounts I manage. This piece sits under how ad production works and what it costs; more about me.

Where you find UGC creators for an AI app depends on what you are buying. For videos to run as ads, the sources are your own team and your own users filming on a phone, a creator you can keep working with, an ad production service, and a marketplace for finished videos (you order a video and one of its creators films it). For people who post from their own accounts, the routes are creator job platforms (where you post a brief and creators apply), the ad platforms’ own creator programs, campus job posts, and searching TikTok and Instagram yourself.

The platform is the second question. The first is which ad you want made, because a creator does better work when you can show them what already works. I run paid ads for subscription apps and games, and I am weaker on organic. I tried a creator job platform once, for my own AI app. Applicants came fast, we took on one creator, and then we stopped.

What are you buying when you ask for UGC?

Different purchases get called UGC, or user-generated content. For an ad, I mean a video filmed by a real person on their own phone that reads as a post. A creator charges for the production of an ad. An influencer charges for reach. So the first thing I ask about a fee is whether it includes posting to the creator’s own audience.

What you buy What you pay for Where people look, by kind My own record
Videos to run as ads Production: a person, a phone, a brief, the raw file and the right to run it Your own team and users, a creator you keep working with, an ad production service, marketplaces for finished videos My daily work, with creators from my network. One brief through TikTok’s creator program for a client, with no result to report.
Creators posting from their own accounts Posts and whatever reach they earn Creator job platforms, the ad platforms’ own creator programs, campus job posts, direct messages Advice more than practice. One listing for my own AI app, with no result to report.
Influencer posts, bought for their audience Access to an audience that already exists Search on TikTok and Instagram, search tools On EatBetter I tested existing creator content as ads and did not recruit the creators. For my own app I prepared outreach twice, with no campaign to report.

The last column is what I have done myself. It is not a ranking.

The only fee I can give you is the creator fee I quote when I sell ad production. When an idea needs a person on camera, that fee comes on top of a round of test ads, typically €100 to €300 or $100 to $300 per creator. It is not a market rate, and a fee per post with a view bonus is a different product.

Do you need a creator yet?

Not always, and often not first. UGC is a format to test inside a creative testing system, and it does not win by default.

When you go to creators, you should already have a good idea of what is working, so they can do something similar or better. The same goes for influencers. Without a few concepts for them to recreate, they are far too expensive. What I want from creators is new concepts that can beat the current ads. If a video also goes viral, that is extra. On my own app I did not have that starting point. We had not had a video go viral ourselves, so I did not know what to tell creators to make.

So I start with the team. If you have a compelling story, someone talking and then showing your app does the job. Ad libraries are a good place to see a category’s top ads, and I would make my own version of the best ones. I look at the long-running ads. What a competitor launched last week is still their test, unless the category runs on trends or your own winners live for months.

Some apps are the exception. For a student app in the US, it may be better to recruit students there through creator networks than to film something yourself.

The app has to be ready as well. The first time I prepared influencer outreach for my own app, I paused it to fix the onboarding first, so that the traffic would not be wasted once it came.

Where would I look, and what happened when I tried?

I describe kinds of platform here. I do not name creator platforms on this site, and I could not rank them from experience anyway.

One team asked me where to find affordable creators for organic reach. I gave them four routes, in no fixed order:

  • campus job posts
  • a listing on a creator job platform
  • searching TikTok or Instagram for people who already post about the topic
  • a platform that pays creators on performance

For my own AI app we put one listing on a creator job platform in October 2026. By our count it passed 100 applicants in under a day. We took on the first creator for three videos. Then we did not take it further. Reaching US viewers this way looked hard. The usual setup, paying several creators a fixed amount each month, did not fit us. At the time I saw one use for such a platform: to find out whether very talented, motivated people who can work on their own are on it.

For a client I have also sent a brief through TikTok’s own creator program. I have no result to report from it.

Two practical points come from my own planning, so treat them as plans. Payments and agreements will come up in creator work, so consider running it on a platform that handles them. And do not wait until you have a pool of creators. Start with the first person you find and see how it goes.

When you buy from a marketplace, the risk I see is footage with no strategy attached. Most creative is briefed across a gap. The people who see the numbers write a document, and the people who make the ads work from it. So ask every platform or partner two things: who briefs the creators, and who feeds the results back.

With a creative partner or a studio, “Can I see the ads you made for other clients?” is the wrong ask. “Can you show me the IPMs those ads actually delivered?” is the better one. IPM is installs per 1,000 impressions.

What is different when the app is an AI app?

Show what the person gets. A shop app sells the shoe. For an AI app, that is the thing the app made for someone. I would still test a version that shows what the app can do, because that is what interests some people. A feature demo may have had a wow effect when these products were new. After years of ChatGPT and a hundred other tools, my reading is that it no longer does. I expect a skeptic of an AI app to think “I can just ChatGPT this”, so I would answer that as early in the ad as I can.

My own example is Subtext, the AI writing assistant I cofounded. It has three UGC ads and one AI presenter ad. A creator I work with is on camera in the UGC ads. Each ad is built around one message people dread sending. I do not publish performance figures for them, so this shows how I applied the idea. It does not prove that the idea works.

My views on AI presenters do not all point the same way. AI footage is cheap enough to put an opening line in front of a real audience before anyone books a shoot, and that is what I use it for. If a line won, I would usually put a real creator on it for the finished ad. The ads that scale on the accounts I run are mostly real people on their own phones.

Against that, creatives made with AI help can outperform handmade ones. I have seen it happen in install rounds, the cheap first test judged on installs. On my own app, though, I called a set of AI-made openings AI slop and asked for more realistic ones. On AI testimonials, I wrote in 2025 that I did not have a strong take yet.

The rules are not my field. What follows is desk research, and it is not legal advice. In the US, the FTC’s rule on fake testimonials covers testimonials by people who do not exist. Meta says it labels ads made or significantly edited with its own AI tools, and that the label sits next to the Sponsored label when the ad shows a photorealistic AI-made person. For ads made with third-party AI tools, Meta says it adds an AI info label when it detects them. TikTok asks advertisers to declare media made or significantly changed by AI. TikTok also says a Spark Ad has to follow its Community Guidelines, because it lives on TikTok as an organic post. In the EU, Article 50 of the AI Act has applied since 2 August 2026, and a realistic invented presenter may count as a deepfake the advertiser has to disclose.

How do I pick a creator?

There are two kinds, and what I would check differs.

For someone who films an ad, you pay for production and not for an audience, so a follower count tells you little. When one team asked, I told them to find the person whose attitude is that they will make it work no matter what. On my own app, my preference was someone who likes the app, uses it and understands it. My working rule with anyone who makes ads for me is one ad before any batch, because you cannot expect 10 ads from someone who has not made one.

For someone you would pay for reach, I wrote a checklist for my own app. I shortlisted with it and no campaign followed, so read it as how I screen and not as a method with a result behind it. At most 100K followers. Look at the median of the last 15 videos and ignore the best one. Check at least five posts, since up to three may be pinned. Look for comments that read like a real community. A tidy profile can be a red flag, because the median looks very good once someone has deleted the videos that did not work. What you can see from outside is gaps in the posting dates.

What do I hand a creator, and what goes in the agreement?

For a video I buy to run as an ad, I handle the direction, the script and the opening line. Usually the person names the problem in their own words before the app appears. One big benefit is enough, two at most. The opening line says why someone needs that benefit. The real app screen comes in early, readable, with subtitles.

For creators who post to their own audience, I wrote my own app’s terms the other way: no scripts and no revision cycles. I wanted the brief to be two to three sentences at most, because nobody reads briefs. One question from that work is still open for me. A video can go viral and not drive downloads or subscriptions, and I do not know how to push for conversion without limiting the creator’s freedom.

This checklist is my own and does not replace legal advice. Each point below is something to agree on and to have covered by the fee.

  • The raw file, because one shoot should become several versions.
  • The right to run the video as a paid ad, and for how long. I agree on the term with each creator and write it into the quote before production.
  • Whether stills, shorter cuts and AI-made versions are covered.
  • Whether you may run the ad from the creator’s handle.
  • That the paid relationship is disclosed in the post.

Keep the agreement short. My own first creator agreement was too long, and my note to myself was to stop being overprepared and get to work.

The creator program I wrote for my own app asks for ad rights at the start, because I cannot tell in advance which video I will want to put spend behind.

I would rather run a creator’s video from the creator’s own handle than upload the file as a new ad. On Meta I set that up as a partnership ad, with an ad code from the creator or with a partnership request. On TikTok it is Spark Ads. When the ad comes from the account of the person in the video, it looks less like an ad. I did not always hold this view. I once argued against switching every ad to partnership ads, but the numbers for 7 days disagreed with me, so I switched. That was one account and one week.

How do I know whether a creator’s video worked?

Test it like any other ad. The example I have published is EatBetter, a health and fitness subscription app, from work that ran in 2025. Its ads came from creator content that already worked organically, and I did not recruit those creators. Each video went through an elimination round. With the paywall optimized alongside, monthly recurring revenue (MRR) went from $10K to $180K in two months, and blended ROAS (return on ad spend) went from 70% to 230% in the same window. Blended means revenue from everyone who installed in a period, paid and organic together, set against that period’s ad spend. I reported these figures for my own client, and nobody audited them. The content worked as paid ads because it already read as a post and was tested like an ad.

One creator should give you several ads. Each version changes one thing, usually the opening. An install round at $20 to $30 a day per test shows which versions get attention. The ad that installs best is not always the ad that keeps people. Purchases decide. Then the winner gets new versions. I keep seeing winners forgotten after they win. My post on a creative testing system for Meta covers how many purchases a version needs before I judge it, and how many versions a budget can judge in a week.

For a paid post, do the sum before you pay. Take the views you expect, a worst and a best case for the share of viewers who install, and your own share of installs that pay. Compare the cost per paying user in each case with what Meta already costs you. Your CPA ceiling, the most a paying user may cost you, is the other number I would check it against. Organic videos do not necessarily bring installs, so an organic hit is still only a candidate for an ad. I would run some posts as ads and see whether the math works.

Sources and scope

For this piece I read the FTC, Meta, TikTok and European Commission pages on 14 November 2026. The EatBetter figures come from my published case page. The fee range and the test budget are my own numbers, and neither is a benchmark. I sell ad production for apps myself, so I have a stake in one of the routes this article describes.

Where my experience stops:

  • Paid ads for subscription apps and games on Meta, Google, TikTok and Apple Ads is what I know. Honestly, I am not super strong at organic.
  • Every creator program I have seen work was run by someone inside the company. Ads and organic are different jobs.
  • My accounts have ad history to read. Without that history, the only evidence is what competitors run, and my rules fit less well.

Sources:

Questions people ask

Where do I find UGC creators for an AI app?

It depends on what you are buying. For videos to run as ads: your own team and your own users filming on a phone, a creator you can keep working with, an ad production service or a marketplace that sells finished videos. For people who post from their own accounts: creator job platforms, the creator programs the ad platforms run themselves, campus job posts and searching TikTok and Instagram yourself. I sell ad production myself.

Which platform is best for finding UGC creators?

I cannot rank them from experience. I put one listing on a creator job platform for my own app, and we stopped after the first creator. I sent one brief through TikTok's creator program for a client. I have no result to report from either. Before any platform, I would ask whether you know which ad you want made, who briefs the creator and who feeds the results back.

Do I need a UGC creator at all?

Not always, and often not first. UGC is one format to test inside a creative testing system, and it does not win by default. A person on the team with a phone and a compelling story can be enough, and creators do better work once you can show them what already works. Some apps are the exception. A student app, for example, may do better with students recruited through a creator network.

What should I agree on with a UGC creator?

Start with one video before any batch. Then agree on the raw file, on the right to run the video as a paid ad and for how long, on whether stills, shorter cuts and AI-made versions are covered, and on whether you may run it from the creator's handle. This is my own checklist, and it does not replace legal advice.

Should an AI app use an AI presenter or a real creator?

I use an AI presenter to find which opening line people stop for, and I would usually put a real creator on the line that won. I have seen ads made with AI help beat handmade ones in install rounds, the cheap first test that is judged on installs. The ads that scale on the accounts I run are mostly real people on their own phones.