A method of generating search queries based on digitized audio from conversations, including: providing a database having a global hot-list of universal popular keywords or phrases and a personalized entity list comprising keywords and phrases used with a frequency above a determined threshold value in conversations involving a user; monitoring a conversation between at least two people, including the user; identifying words or phrases in digitized audio of the monitored conversation through speech recognition; comparing the identified words or phrases to the keywords and phrases in the database to find any matches; generating a search string, without the user requesting a search, based on words or phrases found to match the keyword or phrases stored in the database; submitting the search string to a search engine as a search query; and serving a set of search results returned by the search engine to a display device of the user.