A new selection method for analysing high-fold coincidence data from large gamma-ray arrays, called gamma-ray filtering (GRF), is presented. The method abandons the conventional concepts of gating and gate boundaries and instead calculates merit functions for detected gamma-ray energies to determine the relative value of incrementing them into the desired spectra. Results show that the selection power and limit of sensitivity are increased approximately by a factor of two when using the new filtering method, something that would be very costly to achieve by an improvement in detector hardware. Results show that for future arrays, capable of collecting higher-fold data, the new method will further increase detection sensitivity over current methods, possibly by a factor of four or more. To implement the new method requires only a modest increase in disk storage and CPU-time requirements.