Stop fake reviews before they post.
Yelp removed 13 million fake reviews in 2025, after they were already live. Google Maps removed 170 million. By the time moderation catches up, the rating's already been read, the booking's already been made, the reputation's already been mis-set. Manav refuses the submission at the API.
The big platforms moderate after the fact. By then it's too late.
Five-star and one-star bombs hit a listing within hours of campaign launch. Even when caught, the reputational damage lands first. Manual moderation can't operate at the speed of a single GPT-5o-powered review farm.
One verify() call before the review writes.
// reviews/submit.ts async function submitReview(req, reviewer) { const { is_human, trust_score } = await manav.verify({ email: reviewer.email, context: "reviews/submission" + ":" + req.body.listing_id }); if (!is_human || trust_score < 50) { return { rejected: "requires_human" }; } return reviews.insert(req.body); }
The trust score lets you tune. Default threshold of 50 catches mass-spawn bot networks. Bump to 70 to catch low-history sockpuppet accounts. Bump to 85 if you want only long-history verified humans (premium reviewer programs).
What you save in moderation costs alone.
Industry-average $0.42 per manually-moderated review. Industry-average 12% of submissions need manual review. Manav cuts that to under 2%.
CAPTCHA is dead. Trust score is alive.
Solver farms break CAPTCHA at $0.001 per challenge. They cannot, however, produce a verified-human identity bound to a passkey and a multi-platform history. The cost asymmetry flips. Manav charges $0.005 per verify; defeating a Manav check costs the attacker hundreds of dollars in plausible identity manufacture, per account, per platform. It's not worth it.