Aims. More than half of all active galactic nuclei show strong photoionised outflows. A major uncertainty in models for these outflows is the distance of the gas to the central black hole. We use the results of a massive multiwavelength monitoring campaign on the bright Seyfert 1 galaxy Mrk 509 to constrain the location of the outflow components dominating the soft X-ray band. Methods. Mrk 509 was monitored by XMM-Newton and other satellites in 2009. We have studied the response of the photoionised gas to the changes in the ionising flux produced by the central regions. We used the five discrete ionisation components A–E that we detected in the time-averaged spectrum taken with the RGS instrument. By using the ratio of fluxed EPIC-pn and RGS spectra, we were able to put tight constraints on the variability of the absorbers. Monitoring with the Swift satellite started six weeks before the XMM-Newton observations. This allowed us to use the history of the ionising flux and to develop a model for the time-dependent photoionisation in this source. Results. Components A and B are too weak for variability studies, but the distance for component A is already known from optical imaging of the [O iii] line to be about 3 kpc. During the five weeks of the XMM-Newton observations we found no evidence of changes in the three X-ray dominant ionisation components C, D, and E, despite a huge soft X-ray intensity increase of 60% in the middle of our campaign. This excludes high-density gas close to the black hole. Instead, using our time-dependent modelling, we find that the density is very low, and we derive firm lower limits to the distance of these components. For component D we find evidence for variability on longer time scales by comparing our spectra to archival data taken in 2000 and 2001, yielding an upper limit to the distance. For component E we derive an upper limit to the distance based on the argument that the thickness of the absorbing layer must be less than its distance to the black hole. Combining these results, at the 90% confidence level, component C has a distance of  >70 pc, component D is between 5–33 pc, and component E has a distance  >5 pc but smaller than 21–400 pc, depending upon modelling details. These results are consistent with the upper limits that we derived from the HST/COS observations of our campaign and point to an origin of the dominant, slow (v < 1000 km s-1) outflow components in the NLR or torus-region of Mrk 509.