We present the XMM-Newton follow-up for confirmation of Planck cluster candidates. Twenty-five candidates have been observed to date using snapshot (~10ks) exposures, ten as part of a pilot programme to sample a low range of signal-to-noise ratios (4 < S/N < 6), and a further 15 in a programme to observe a sample of S/N > 5 candidates. The sensitivity and spatial resolution of XMM-Newton allows unambiguous discrimination between clusters and false candidates. The 4 false candidates have S/N ≤ 4.1. A total of 21 candidates are confirmed as extended X-ray sources. Seventeen are single clusters, the majority of which are found to have highly irregular and disturbed morphologies (about ~70%). The remaining four sources are multiple systems, including the unexpected discovery of a supercluster at z = 0.45. For 20 sources we are able to derive a redshift estimate from the X-ray Fe K line (albeit of variable quality). The new clusters span the redshift range 0.09 ≲ z ≲ 0.54, with a median redshift of z ~ 0.37. A first determination is made of their X-ray properties including the characteristic size, which is used to improve the estimate of the SZ Compton parameter, Y500. The follow-up validation programme has helped to optimise the Planck candidate selection process. It has also provided a preview of the X-ray properties of these newly-discovered clusters, allowing comparison with their SZ properties, and to the X-ray and SZ properties of known clusters observed in the Planck survey. Our results suggest that Planck may have started to reveal a non-negligible population of massive dynamically perturbed objects that is under-represented in X-ray surveys. However, despite their particular properties, these new clusters appear to follow the Y500-YX relation established for X-ray selected objects, where YX is the product of the gas mass and temperature. Corresponding author: E. Pointecouteau, e-mail: etienne.pointecouteau [at] irap.omp.eu (etienne[dot]pointecouteau[at]irap[dot]omp[dot]eu)