Notes. For periods longer than 200 days, these values could be underestimated, as explained in the text. Error bars give 1-σ confidence intervals, but a value of ϵ < 0.5 mas leads to an upper-limit on mass and semi-major axis and a lower limit on inclination (see text). Only the sources for which the uncertainty on the period is better than 10% are shown. For triple systems, we calculate the inclination for companion b orbit, assuming that only the inner short-period companion has a measurable effect on Gaia astrometry. We marked with an asterisk the BD candidates whose true mass is estimated to be above 90 M_J at 1σ. The definition of the marginal probability is explained in Sect. 6.1. The minimum inclination I_c,min measures the minimum inclination a system can have while the secondary companion remains dark. ^(a) The likelihood of ϵ ± 10% given that I_c > I_c,− (or I_c < I_c,+) can be calculated by using formula (12) and (13), replacing p by 0.683, p(ϵ) by the marginal probability, and I_c,min by the value given in this column. ^(b)The marginal probability as defined in Sect. 6.1.3. ^(c) These targets were not part of the TGAS sample. Therefore, ϵ_DR1 measures the scatter around the five-parameter solution only accounting for the 14 months of Gaia measurements. In these cases, the proper motion is fitted and modelled out of the simulations. ^(d) For HD 71827, ϵ is larger than any of the simulations that assume a dark companion, i.e. M_V,2 − M_V,1 > 2.5. The excess noise observed by Gaia is likely due to the large 20-yrs period outer companion (see text explanation in Sect. 6.1.4). ^(e) For HD 62923, ϵ is at the very limit of simulations that assume a dark companion, i.e. M_V,2 − M_V,1 > 2.5. In this system, the secondary likely pollutes the astrometry as emphasised in Sect. 5.