Web pages of a website are parsed and a set of n-grams are generated from the parsed web pages. A relevancy value is determined for each n-gram and a second set of n-grams is generated by removing any n-gram in the first set whose relevancy value is below a threshold. A third set of n-grams is generated at least by removing those of the second set of n-grams that have been determined to be similar to another one of the second set of n-grams. Responsive to determining that there is not a web page that is directed at an n-gram, a web page is automatically created with content directed at that n-gram including reusing existing content of the website that is related to the n-gram. One or more links to the created page are added to web pages so that the created page is not an orphan page.