What's useful
- Interaction logs: prompts, outputs and multi-step agent traces.
- Feedback signals: ratings, edits, retries and accepted or rejected suggestions.
- Task outcomes: whether the user's goal was actually achieved.
- Evaluation sets you've built to test your own product.
What qualifies
The listing is for AI product companies of roughly 5 to 200 employees with a live product and real users.
What to check first
- Your terms and privacy policy. Licensing usage data to a third party must be something your terms of service and privacy notice actually allow. If they don't, you'd need to change them going forward, and that may not cover data you already hold.
- Business customers' contracts. Enterprise and team plans often promise that customer content won't be used for anything else. Those customers' data is usually out of scope.
- Personal and confidential content. Prompts often contain personal data, and sometimes other companies' confidential information or code. Ask how it would be filtered.
- Upstream model terms. Check whether the terms of any model provider you use restrict how outputs can be used.
What it's worth
Mercor lists potential company earnings of $20kâ$480k for AI product companies, depending on what's in scope. That is Mercor's own figure, not a promise or guarantee from us.
Can we include only users who opted in?
Yes, and that's often the simplest basis. Say how many opted-in users you have when you get in touch.
We're early stage. Is our data enough?
Volume matters, but so does quality. Real usage with clear outcomes can be useful even at modest scale.