Prolific powers personality-driven AI evaluation
$350M/year Mercor for human personality
This is a bet that the scarce input in AI is shifting from raw intelligence to human fit. Mercor and Handshake win when a lab needs a doctor, lawyer, or physicist to grade answers. Prolific wins when a builder needs calm people for support flows, culturally fluent users for localization, or politically varied respondents to test how an assistant lands across real audiences.
-
Prolific’s edge is not just a big worker pool, but years of profiling and repeat performance data on roughly 200,000 active participants, plus millions more on the waitlist. That lets customers filter for traits like resilience, language fluency, geography, experience, and behavior, then fill studies in hours or minutes.
-
The workflow is concrete. An AI company can use the audience discovery tool or API to find, for example, middle aged bilingual users in a specific country, or people who stay calm under pressure, then have them red team prompts, rate bedside manner, test tutoring quality, or compare support conversations.
-
This is the next step after anonymous crowd labor and expert labor. Scale grew on broad annotation, then Mercor and Handshake grew on verified experts. Prolific is pushing toward tasks where the right answer depends on how a person feels, reacts, trusts, or reads social context, which is exactly what assistant and companion products need.
The market is heading toward always on human evaluation for live products, not just one time model training. As assistants move into therapy, education, healthcare, and customer support, platforms that can reliably supply specific kinds of people, not just smart people, should capture more of the spend and become core infrastructure for post training, safety, and product tuning.