Methods and apparatus to meter content exposure using closed caption information are disclosed. An example method comprises developing a keyword database of terms based on program guide descriptive of programs for a given time period, generating one or more values representative of likelihoods that one or more respective media content was presented based on a comparison of closed caption text and the keyword database, collecting audience measurement data, and employing the one or more likelihood values to identify a set of reference data for comparison to the audience measurement data to identify presented content.