Strong He II emission is produced by low-metallicity stellar populations. Here, we aim to identify and study a sample of He II $\lambda 1640$-emitting galaxies at redshifts of $z \sim 2.5-5$ in the deep VANDELS spectroscopic survey.. We identified a total of 33 Bright He II emitters (S/N > 2.5) and 17 Faint emitters (S/N < 2.5) in the VANDELS survey and used the available deep multi-wavelength data to study their physical properties. After identifying seven potential AGNs in our sample and discarding them from further analysis, we divided the sample of \emph{Bright} emitters into 20 \emph{Narrow} (FWHM < 1000 km s$^{-1}$) and 6 \emph{Broad} (FWHM > 1000 km s$^{-1}$) He II emitters. We created stacks of Faint, Narrow, and Broad emitters and measured other rest-frame UV lines such as O III] and C III] in both individual galaxies and stacks. We then compared the UV line ratios with the output of stellar population-synthesis models to study the ionising properties of He II emitters. We do not see a significant difference between the stellar masses, star-formation rates, and rest-frame UV magnitudes of galaxies with He II and no He II emission. The stellar population models reproduce the observed UV line ratios from metals in a consistent manner, however they under-predict the total number of \heii ionising photons, confirming earlier studies and suggesting that additional mechanisms capable of producing He II are needed, such as X-ray binaries or stripped stars. The models favour subsolar metallicities ($\sim0.1Z_\odot$) and young stellar ages ($10^6 - 10^7$ years) for the He II emitters. However, the metallicity measured for He II emitters is comparable to that of non-He II emitters at similar redshifts. We argue that galaxies with He II emission may have undergone a recent star-formation event, or may be powered by additional sources of He II ionisation.