We discuss the exclusive photoproduction of a photon-meson pair with large $ p_{T} $ as a channel to probe generalised parton distributions (GPDs). Like other $ 2\to 3 $ exclusive processes, this channel allows us to better study the $ x $-dependence of GPDs, in contrast to $ 2\to 2 $ processes such as Deeply Virtual Compton Scattering (DVCS) which give only ``moment-type'' information. Moreover, it also gives the possibility to access, at leading twist, the chiral-odd GPDs, which are still completely unknown experimentally. In the first part of these proceedings, we present the computation of the amplitude at leading order and leading twist, for charged pions, and rho mesons of any charge and polarisation. These processes are sensitive to quark GPDs only. We also discuss the possibility of measuring such processes at various experiments, namely JLab, COMPASS, future EIC and LHC in ultra-peripheral collisions. In particular, in collider experiments, the dependence of GPDs at small skewness $ (<10^{-3}) $ can be extracted. For the second part, we report on our recent work on the exclusive photoproduction of $ \pi^{0}\gamma $ pair, which, in addition to a quark GPD channel, also has a contribution from gluon GPDs. In the latter case, we demonstrate that a collinear factorisation of the process fails, due to the presence of a Glauber pinch. Our findings also imply that other similar processes, like $ \pi ^{0} N \to \gamma \gamma N $ also suffer from the same issue. On the other hand, we stress that the processes discussed earlier, which are sensitive to the quark GPD channels only, are safe from these factorisation breaking effects.