Fermi level can be defined for semiconductors as an ideal level at which the probability of occupation with electrons is 50% which lies in the band gap at (T=0K) . In metals, it’s defined as the top occupied energy level in the conduction band at (T=0K). Thus we have two definitions, one for semiconductors and the other for metals.
Why we don’t try to unify the definitions?
I mean let us go deep in the Fermi Dirac distribution function, the 50% occupation rule is obvious. What about the second definition of metal?
Metal's conduction band is practically a continuous energy band, i. e. energy levels are very closely spaced such that they can practically be considered continuous (I found it in 'Principle of Solid State Physics' by Levy). Now what happens if we say that the first definition is also working in metals?
The top occupied level in metals is very close to the above unoccupied level (they are almost overlapping). Now, the ideal energy level which lies between the top occupied level and the lower unoccupied level is practically the top occupied level. Why we don't say that the first definition leads to the second and vice versa?