A method, performed by one or more server devices, includes receiving an abuse report from a user, where the abuse report indicates that a comment associated with a document violates a policy associated with providing comments to users; retrieving a trust score of the user, where the trust score is based on a difference between a quantity of valid abuse reports previously generated by the user and a quantity of invalid abuse reports previously generated by the user; generating or adjusting a removal score of the comment based on the retrieved trust score of the user; comparing the removal score to a removal threshold; and removing the comment from further presentation to users if the removal score of the comment is greater than the removal threshold.