Disclosed are an apparatus and a method for controlling a tracking servo in a near field optics. Gap error signals are detected in a divided state, and a push pull signal is compensated based on the detected gap error signals. A photo detector detecting the gap error signals is divided into two areas such that movement of an image formed on the photo detector can be detected. The push pull signal is compensated by subtracting a value proportional to a difference value between the two gap error signals detected in the photo detector. The difference value between the two gap error signals is proportional to a dc offset component included in the push pull signal. Accordingly, it is possible to simply remove a dc offset included in a push pull signal without largely modifying the near field optics, so that it is possible to exactly perform a tracking servo operation.