As wind power is growing towards a major utility source of electricity, reducing the cost of energy becomes a critical issue in order to make wind power competitive to conventional sources. The current trend in wind turbine (WT) development is for larger, more complex machines located in remote offshore locations which present operation and maintenance (O&M) challenges and high costs. Significant savings can be made by establishing technically and economically viable condition-based maintenance (CBM) through the implementation of automated condition monitoring system (CMS) data interpretation. Gearbox faults, with high replacement costs, complex repair procedures and revenue loss due to long downtime, are widely considered as the leading issue for WT drive train condition monitoring. The purpose of this paper is to design a fault detection algorithm for incorporation into a commercial CMS for automatic gear fault detection and diagnosis. To this aim, experimental work was carried out on a condition monitoring test rig equipped with a commercial WT CMS. Gear tooth damage has been investigated by establishing pattern feature parameters from gearbox vibration signatures. The results demonstrate that the presence of meshing frequency harmonic sidebands and their amplitudes can prove to be very valuable when diagnosing gear defects. A gear fault diagnostic algorithm, able to track the overall power of the spectra sideband frequency window, was proposed in order to reduce each spectrum to only one parameter and automate the data interpretation. The algorithm has proved successful to detect the progression of a gear tooth defect even at the early development stages.