A method for detecting Web sites used for phishing, including preselecting one or more Web sites to be examined for duplication, selecting at least one or more elements that are present in the preselected Web site and that relate to characteristic identifying features of the preselected Web site, forming at least one search query using the one or more elements, and submitting the at least one search query to an indexed public search engine. The elements illustratively may be URL substrings, content identification substrings, or tree structure-related substrings. A report of Web sites using the selected one or more search terms is received from the public search engine in response to the query, and the preselected Web site is eliminated from the Web sites found in the search. The remaining Web sites retrieved in the search are further analyzed, by additional focused searching of the retrieved pages, by comparing header or tree structure information, or other techniques to compare them with the preselected Web site to identify unauthorized near-replicas of the known legitimate Web site for responsive action.