ABSTRACT For sensitive optical interferometry, it is crucial to control the evolution of the optical path difference (OPD) of the wavefront between the individual telescopes of the array. The OPD between a pair of telescopes is induced by differential optical properties such as atmospheric refraction, telescope alignment, etc. This has classically been measured using a fringe tracker that provides corrections to a piston actuator to account for this difference. An auxiliary method, known as the Piston Reconstruction Experiment (P-REx), has been developed to measure the OPD, or differential ‘piston’ of the wavefront, induced by the atmosphere at each telescope. Previously, this method was outlined and results obtained from Large Binocular Telescope adaptive optics data for a single telescope aperture were presented. P-REx has now been applied off-line to previously acquired Very Large Telescope Intereferometer (VLTI)’s GRAVITY Coudé Infrared Adaptive Optics wavefront sensing data to estimate the atmospheric OPD for the six baselines. Comparisons with the OPD obtained from the VLTI GRAVITY fringe tracker were made. The results indicate that the telescope and instrumental noise of the combined VLTI and GRAVITY systems dominates over the atmospheric turbulence contributions. However, good agreement between simulated and on-sky P-REx data indicates that if the telescope and instrumental noise was reduced to atmospheric piston noise levels, P-REx has the potential to reduce the OPD root mean square of piston turbulence by up to a factor of 10 for frequencies down to 1 Hz. In such conditions, P-REx will assist in pushing the sensitivity limits of optical fringe tracking with long baseline interferometers.