Aims. We investigate the asteroseismology of two solar-like targets as observed with the CoRoT satellite, with particular attention paid to the mode fitting. HD 181420 and HD 49933 are typical CoRoT solar-like targets (156 and 60-day runs). The low signal-to-noise ratio (SNR) of about $3{-}10$ prevents us from unambiguously identifying the individual oscillation modes. In particular, convergence problems appear at the edges of the oscillation spectrum. Methods. We apply a Bayesian approach to the analysis of these data. We compare the global fitting of the power spectra obtained by the classical maximum likelihood (MLE) and the maximum a posteriori (MAP) estimators. Results. We examine the impact of the choice of the priors upon the fitted parameters. We also propose to reduce the number of free parameters in the fitting, by replacing the individual estimate of mode height associated with each overtone by a continuous function of frequency (Gaussian profile). Conclusions. The MAP appears as a powerful tool to constrain the global fits, but it must be used carefully and only with reliable priors. The mode width of the stars increases with the frequency over all the oscillation spectrum.