The crystal structures of TiIII
4[Si2O(PO4)6] (
P
3
‾
$P\overline{3}$
, Z = 3, a = 14.733(1), c = 7.363(1) Å, R1 = 0.040, wR2 = 0.098, 7649 ind. refl., 170 variables), FeII
0.79TiIII
2.42TiIV
0.79[Si2O(PO4)6] (
P
3
‾
$P\overline{3}$
, Z = 3, a = 14.6534(2), c = 7.3829(1) Å, R1 = 0.036, wR2 = 0.088, 4026 ind. refl., 171 variables), and TiIII
2TiIV
6(PO4)6[Si2O(PO4)6] (
R
3
‾
$R\overline{3}$
, Z = 1, a = 8.446(2), c = 44.21(2) Å, R1 = 0.047, wR2 = 0.120, 1373 ind. refl., 109 variables) have been refined from single-crystal data. The structures show hexagonal closest packing of phosphate groups with metal cations and [Si2O] groups occupying octahedral voids [□(PO4)6]. The close relationship of these and other silicophosphate structures to the NiAs and β-Fe2(SO4)3 (see also NaZr2(PO4)3 “NASICON”) structure types is rationalized by group/subgroup considerations. This symmetry approach shows that systematic twinning is highly likely in silicophosphates, thus possibly leading to faulty crystal structure refinements. Our investigation strongly suggests that the proper composition of silicophosphates “M
III
3P5SiO19” (M = Cr, V, Fe, Mo) reported in literature is actually M
III
4-[Si2O(PO4)6]. In the mixed-valent compounds oxidation states were assigned to the cation sites by comparison to Ti2O3, TiP2O7 and FeTiO3. The powder reflectance spectrum of dark-blue FeII
0.79TiIII
2.42TiIV
0.79[Si2O(PO4)6] shows a strong IVCT transition at
ν
˜
$\widetilde{\nu }$
= 17,500 cm−1, and magnetic susceptibility data agree very well with the proposed oxidation states.