The companies building AI models need an enormous amount of content to train on. The problem, until recently, was that creators had no clean way to be found by those buyers, no way to set a price, and no way to get paid without a licensing negotiation that takes months.
JubJub changes all three. If your content is published on JubJub, it is already discoverable, already priced, and already purchasable by AI buyers. You do not need to do anything extra. The infrastructure is already working for you.
01The marketA market that did not exist two years ago.
AI training data is now a real commercial category, and it is growing fast. Model developers need large volumes of varied, high-quality human-made media. The content they most want is the content that is hardest to find at scale: genuinely human-captured, clearly attributed, and verifiably not AI-generated.
The problem for creators has always been the same: how do you get in front of buyers who are moving at machine speed, how do you protect your work from being scraped for free, and how do you get paid in a way that does not require a legal team?
JubJub solves all three. Your content is discoverable via the MCP server, priced by you, and purchasable only via x402 or MPP. No scraping path. No free access. No middleman taking the lion’s share.
02What buyers wantWhat AI buyers actually want from your content.
When an AI company or autonomous agent evaluates content for training use, it needs answers fast. The content either fits the task or it does not, and the agent needs to know before it commits to buying. The questions it is asking:
- Is this human-generated or AI-generated?
- What subjects, languages, and interaction types does it cover?
- Is there a transcript and visual frame reference for alignment work?
- What is the provenance record?
JubJub’s media intelligence payload answers all of them at the point of purchase. The agent receives a full transcript, reference frames, and structured metadata. It gets a complete picture of what your content contains before it commits to buying the full item.
You stop leaving money on the table. That is the only thing that changes.
03The moneyHow the money actually works.
Per-use purchasing
You set a price per content item at publish time. An agent queries the catalogue, requests a sample, decides it wants the full payload, and pays.
- USDC settles directly to your wallet
- JubJub takes 3%
- No payout hold
Bulk dataset licensing
Offer your full catalogue or a curated subset as a structured licensed dataset. If that model makes sense for your volume of content, it is coming.
- Full catalogue or curated subset
- Structured, licensed dataset
- For now, per-use is how it works
04Still just publishingYou are not becoming a data vendor.
This is worth saying clearly. You are not launching a side business. You are not building a data catalogue or managing licensing contracts. You are publishing content the same way you always have.
What JubJub does is make that content available to a second type of buyer running in parallel to your human audience. If an AI agent finds your work valuable for its task, it pays. If it does not, nothing changes. Your human revenue streams are unaffected. Your workflow is unaffected.
You stop leaving money on the table. Same content, same workflow, same human audience. A second buyer pool simply starts paying for work you were already publishing.
05Why origin mattersWhy the origin of your content is starting to matter to buyers.
AI companies are increasingly aware that the quality of a model depends on the quality of what it trains on. Content that is genuinely human-captured, clearly attributed, and verifiable is worth more than content of unknown origin. It produces better models. Buyers are starting to price that in.
Most creators cannot prove their content meets that bar. They can say it does. They cannot verify it in a form an AI system can read and act on. JubJub’s ownership infrastructure creates that verification record automatically at point of publication. Your content arrives with the record already attached.
This is a thesis the market is still catching up to. But the direction is clear, and being early here is a real advantage.
