The Smartest Creators Are Now Licensing Their Content as Training Data

Lucas: So I was on the phone last week with a food creator based in Austin — she has about 900,000 followers on TikTok, mostly recipe videos. And she told me about a check she got in April that I had to ask her to repeat twice. Luna: I'm guessing it wasn't from brand deals or the Creator Fund. Lucas: Not even close. It was for three hundred and forty thousand dollars. One payment. From a company you've heard of — one of the big large language model players. And it wasn't for ad space, it wasn't for sponsorship, it wasn't for a course. She licensed them access to her video library as training data. Luna: Wait — they paid her three hundred and forty grand to train their AI on her cooking videos? Lucas: Exactly. And she's not alone. I've now talked to five creators who have signed similar deals in the past six months. The going rate seems to be somewhere between fifty thousand and half a million dollars, depending on the size of the library, the niche, and — crucially — whether it's exclusive or non-exclusive. Luna: And hers was non-exclusive, you said? Lucas: Non-exclusive. Which means she still owns all the rights, she can still license the same footage to another company, she can still use it herself. The AI company gets a perpetual, non-exclusive license to use that content as training material. That's the model that's really taking off right now. Luna: I want to unpack the numbers. She has about twelve hundred videos, right? So that's roughly two hundred and eighty-three dollars per video. That's a pretty good rate for content you already made and posted. Lucas: It's extraordinary when you frame it that way. And consider that most of those videos were originally monetized through the TikTok Creator Fund — which for a video with a million views might pay you twenty, thirty dollars. Or through brand deals, which are lumpy and require a lot of work. This is pure passive revenue on existing assets. Luna: But who is actually buying this? I mean, we know the big AI labs need data, but I thought most of them had already scraped everything publicly available. Lucas: That's the key insight. They did scrape the public web. But the legal landscape has shifted. In 2024 and 2025, a series of lawsuits — from The New York Times, from Getty Images, from individual creators — made it very clear that using copyrighted content without permission for training is risky. So starting late last year, the major labs began proactively licensing high-quality, curated datasets directly from creators. And TikTok video is uniquely valuable because it's short-form, it's multimodal — visual, audio, text overlays — and it captures real human behavior and expression in a way that scripted TV or static images don't. Luna: So they're not just buying any video. They're buying videos that have high engagement metrics, clear niche focus, and ideally a consistent style. Lucas: Bingo. The creator I mentioned — her videos are all shot in the same kitchen, same lighting, same overhead angle. She has a very consistent plating style. That's gold for training, because the model can learn patterns from a coherent dataset. The companies are basically looking for creators who are mini production houses with a recognizable aesthetic. Luna: If these marketing conversations have sparked something you've actually used — a strategy you tried, a new way of thinking about your content — we just want to say that this show stays ad-free because listeners like you choose to support it. There's no pressure at all, but if you'd like to help keep these conversations going without any sponsor interruptions, the link is buy me a coffee dot com slash fexingo. That's it. No perks, no tiers, just a simple way to say this matters to you. Lucas: Yeah, and it genuinely makes a difference. Every contribution goes straight into keeping the show independent and focused on what's actually happening in the creator economy. Okay — back to those data deals. One thing I find fascinating is that a whole ecosystem of middlemen has sprung up in the last eighteen months. There are now agencies that specialize in representing creators for training data licensing. Luna: Like talent agents, but for your video library instead of your face? Lucas: Exactly. One of them, I spoke with a co-founder — their firm represents about 200 creators. They negotiate the contracts, handle the legal work, and take a 15 to 20 percent cut. Their argument is that a creator going solo would never get the same terms because they don't know what to ask for. For example, one big point of negotiation is the duration of the license. Perpetual is standard, but some creators are pushing for limited-time licenses — three years, five years — so they can re-license later if the market matures. Luna: Are any creators forming collectives to bargain together? Lucas: Yes, and that's actually the most interesting development. There's a group of about fifty travel and lifestyle creators who banded together in March to offer their combined libraries as a single dataset. They're asking for a minimum of two million dollars for a non-exclusive license to the full bundle. That's about forty thousand per creator on average, and they retain individual rights to do separate deals. The leverage comes from offering a curated, diverse set of high-quality content in one transaction. Luna: That's smart. It also solves a problem for the AI companies, which is that they don't want to negotiate a thousand individual licenses. They'd rather write one big check and get a clean dataset with clear rights. Lucas: Exactly. And the collective is using a blockchain-based registry to track usage and enforce terms — which sounds a bit buzzwordy, but in practice it just means every video gets a unique identifier and the license terms are publicly verifiable. I think this is going to become a standard model for mid-tier creators who don't have the individual clout of a MrBeast or a Charli D'Amelio but collectively represent enormous value. Luna: Let's talk ethics for a minute. There's a real tension here. On one hand, creators finally getting paid for the value their content provides to AI systems. On the other hand, some people argue that licensing your content to train AI is essentially selling the rope that will hang you — because eventually these models will be able to generate content that competes with the original creator. Lucas: I've heard that argument, and I think it's valid in certain cases. If you're a creator whose entire value proposition is a specific visual style or a specific type of tutorial, then yes — you might be training your replacement. But the creators I've talked to are pretty clear-eyed about it. They see it as an additional revenue stream, not a replacement for their primary business. The food creator I mentioned — she still does brand deals, she still sells a cookbook, she still runs a paid newsletter. The training data license is maybe 20 percent of her annual income. And she's betting that her personal brand and her live interaction with her audience can't be replicated by a model. Luna: So it's a hedge. Take the money now, because the market might not be this generous later. Lucas: Exactly. And there's also a first-mover advantage. The companies that are buying right now are paying a premium for clean, high-quality data because they're in an arms race. In two years, the market might be saturated, or the legal framework might change, and the prices could drop. So creators who act now are capturing a moment. Luna: What about the smaller creators? Someone with 10,000 followers and a hundred videos — is there any path for them? Lucas: It's harder but not impossible. The agencies I mentioned are starting to aggregate smaller creators too. One agency told me they'll take on creators with as few as 50,000 followers if the content is niche enough — say, a very specific craft like leatherworking or a hyperlocal food scene. The key is uniqueness. If your content is generic — vlogs, reaction videos, trends — it's less valuable because the AI can already find that data. But if you have a very specific skill or perspective, there's a market. Luna: So what's the practical advice for a creator listening right now who thinks their library might be valuable? Lucas: First, inventory your content. Catalog every video you have, with metadata: topic, length, engagement rate, date posted. Second, make sure you have clear ownership — if you used licensed music or stock footage, those clips might not be licensable. Third, reach out to a specialized agency or look into a creator collective. Going solo is tough because the legal terms are complex. And fourth, don't sign an exclusive deal unless the premium is enormous. Non-exclusive gives you flexibility. Luna: One more thing — we're seeing this in video, but is it happening with audio or images too? Lucas: Absolutely. There are podcasters licensing their audio libraries, photographers licensing their image catalogs. But video is the hottest right now because it's the richest data type and the hardest to generate from scratch. TikTok specifically is the holy grail because of the sheer volume and diversity of real-world scenarios captured. Luna: I wonder if we'll look back on 2026 as the year the creator economy got a second revenue pillar — the first being audience monetization, and now data licensing. Lucas: That's exactly what I think is happening. The creator economy has always been fragile — one algorithm change, one brand budget cut, and income drops. Data licensing is a much more stable, long-term revenue stream because it's tied to the structural demand for training data. And that demand isn't going away. If anything, it's going to grow as more industries deploy AI. Luna: Alright, I'm going to go catalog my old travel videos just in case. Thanks Lucas. Lucas: Good luck. If you find a leatherworking video in there, you might be in business.