A computerized method of automatically determining a user's potential liking for an unknown audio file, based upon a reference audio file that the user likes. Each audio file is analyzed into various frequency and amplitude bins, each bin having a time length that is at least a substantial fraction of the audio file's normal playing time. Each bin is further scored according to the complexity or information content of the audio signal in that bin. Corresponding bins from the reference audio file and one or more unknown audio files are then compared on a per-bin basis, and the results aggregated according to an anticipated user affinity score using an anticipated user affinity algorithm. The method can be used by an internet server to accept song input from users (using user computerized devices), and in turn suggest other high user affinity score songs that the user may like.