A method for detecting click fraud. In one embodiment, the method includes the steps of: directing a user to an item but delaying the internal ratings update of the item; collecting data to build a model of the user; building the model of the user in response to the collected data; determining the trustworthiness of the user in response to the model of the user; and updating internal ratings of the item in response to the trustworthiness of the user. In various embodiments, the click fraud can be the inflation of advertisement clicks, the inflation of popularity or the inflation of reputation.