A trust layer for AI agents is overdue, and structured reviews from verified users is the right call. How do you stop agent builders from gaming their own ratings?
Fair question, it's the first problem any review platform has to answer. Right now reviews require a verified email before they go live, and every review passes through moderation before it's public, so nothing goes up unread. The ranking isn't a simple star average either, it's weighted by review volume, so a builder posting 10 five star reviews from friends doesn't move the needle the way it would on a raw average. Builders can claim their listing and respond to reviews, but they can't edit or delete them. And the AI generated assessment is kept separate from user reviews and labeled as such, so there's nothing to game there by stuffing reviews.
Is it bulletproof? No. Email verification alone won't stop someone determined to create multiple accounts, and that's the next problem to solve, probably through account age, usage signals, or requiring more than just an email. But moderation and volume weighted scoring are real deterrents today, not just a policy statement.