Current CAPTCHAs require users to solve objective questions such as text recognition or image recognition. We propose a class of CAPTCHAs based on collaborative filtering. Collaborative filtering CAPTCHAs allow us to ask questions that have no absolute answer; instead, the CAPTCHAs are graded by comparison to other people’s answers. We analyze the security requirements of collaborative filtering CAPTCHAs and find that although they are not ready to use now, collaborative filtering CAPTCHAs are worthy of further investigation.