This is an open question. Metal transition atoms dilutes in the non magnetic semiconductor and can produce several effects, as (between others):
1. Magnetic clusters formation: This is the case of Mn-doped GaAs
2. Inhomogeneous distribution
3. Spinodal decomposition
4. Segregation
The first effect give place to a spontaneous ferromagnetic component (SFC).The second one also give place to SFC sometimes together with a paramagnetic component. The third case produce two phases with the same crystal structure and very close lattice parameters; one phase is rich in the metal element and the other is poor. The fourth case generally produce binary magnetic phases that can be considered as secondary phases.
A very interesting review has been published in 2015 for Dietl et al ( included in attach)
Semiconductor with a small (diluted) concentration of "magnetic" impurity (typically a transition metal) is ferromagnetic if two conditions are satisfied:
1. The Impurities have a large enough magnetic moment in the semiconductor lattice
2. There is a long-range magnetic ordering which aligns the magnetic moments in parallel. For a practical application magnetic order has be strong enough to "survive" thermal fluctuations at high enough temperatures, e.g. Room temperature. There can different mechanisms responsible for ferromagnetic ordering.
For a detailed discussions on Magnetic properties in dilute magnetic semiconductors i suggest you to read the review paper: Rev. Mod. Phys. 82, 1633 – Published 20 May 2010. the physics in III-V materials is more complex than what can often been found in the literature, mean field theories (Zener Mean field) that always overestimate the Curie temperature, neglect the physics of disorder (localisation)and percolation. In Zener Mean field theory, the magnetic curling are RKKY like, which is not the case beyond these drastic approximations. in dilute matrix RKKy couplings leads to low critical temperatures, small region of stability because of frustrations effects (see Phys. Rev. B 73, 024411) ..
Typically below 1 an 2% in Mn doped GaAs the absence of percolation leads to no ferromagnetic long range order, thus Tc=0 (See for example EPL, 92 (2010) 47006)
A lot has been done , but you will se that there is still a lot of confusion in the literature .