The behaviour of gravitational waves in the anomaly-induced inflationary phase is studied. The metric perturbations exhibit a stable behaviour, with a very moderate growth in the amplitude of the waves. The spectral indice is computed, revealing an almost flat spectrum.The goal of the present study is to analyze the fate of gravitational waves in a background defined by an inflationary scenario created by a trace anomaly induced by quantum effects generated by matter fields in the primordial Universe [ 1]. This model is a generalization of the Starobinsky model [ 2], where a similar mechanism has been developed. The counterterms necessary to avoid divergences in the quantization of these conformal quantum fields in a FriedmannRobertson-Walker background leads to a higher derivative action with a trace anomaly. Details of the model may be found in [ 1]. An important point concerns the fact that an inflationary phase may be obtained with this effective gravitational action. When all fields are massless, this inflationary phase is an eternal de Sitter phase, where many of the traditional problems of the inflationary scenario appear. However, if massive conformal fields are included, the inflationary phase has more complicated form. A detailed analysis of the massive case reveals that a huge amplification of the scale factor may be obtained, of the order of 10 10 e-folds. However, from the observational point of view, only the last 65 − 70 e-folds are relevant. During this final period, the behaviour of the scale factor may be approximated by an exponential function, the Hubble parameter being constant. It is possible that a transition to the FRW standard scenario may be achieved solving the graceful exit problem [ 4].The equation governing the behaviour of gravitational waves in the anomaly-induced model, in terms of the cosmic time t, is [ 3]withHere the quantities with tildes are pure numbers:whileb 4 depends on the multiplet content of the matter fields:where N 0 , N 1/2 and N 1 are the scalar, fermionic and vectorial numbers of matter fields. For example, in the minimal standard model, N 0 = 4, N 1/2 = 24 and N 1 = 12, leading to b 4 ∼ 0.2. As usual, H =ȧ/a is the Hubble parameter. The parameters described above take those value in the last 65 e-fold. In this case, the background model enters in a quasi-de Sitter and the scale factor behaves essentially as a ∼ e Ht (5)