Because massive, low-metallicity population III (PopIII) stars may produce very powerful long gamma-ray bursts (LGRBs), high-redshift GRB observations could probe the properties of the first stars. We analyze the correlation between early PopIII stars and LGRBs by using cosmological N-body/hydrodynamical simulations, which include detailed chemical evolution, cooling, star formation, feedback effects and the transition between PopIII and more standard population I/II (PopII/I) stars. From the Swift observed rate of LGRBs, we estimate the fraction of black holes that will produce a GRB from PopII/I stars to be in the range 0.028<f_{GRB}<0.140, depending on the assumed upper metallicity of the progenitor. Assuming that as of today no GRB event has been associated to a PopIII star, we estimate the upper limit for the fraction of LGRBs produced by PopIII stars to be in the range 0.006<f_{GRB}<0.022. When we apply a detection threshold compatible with the BAT instrument, we find that the expected fraction of PopIII GRBs (GRB3) is ~10% of the full LGRB population at z>6, becoming as high has 40% at z>10. Finally, we study the properties of the galaxies hosting our sample of GRB3. We find that the average metallicity of the galaxies hosting a GRB3 is typically higher than the critical metallicity used to select the PopIII stars, due to the efficiency in polluting the gas above such low values. We also find that the highest probability of finding a GRB3 is within galaxies with a stellar mass <10^7 Msun, independently from the redshift.