It's a hard question to answer, since in some cases semi-emperical methods can be just as good as ab initio methods, but in general, semi-empirical tends to be less accurate. Semi-empirical methods are usually only recommended for big systems with a large number of atoms, since it has a lower computational cost, but for small systems (less than 100 atoms), ab initio is usually the best choice.

You must consider electrons in the TS search, hence semi-emperical methods are not reasonable choices.

I modelled a resonable TS guess for my calculation. But on processing even at HF as well as MO6 levels of theories only certain H bonded elongations were exhibited continously. How would one explain this situation?

If a QM calculation is affordable, you should really attempt to characterize your transition state vibrationally.

Semiempirical methods are based on heavy parametrization and usually lack flexibility when describing weird systems, such as transition states. Hydrogen atoms are particullarly difficult to model in awkward situations, for example in displacements that take them far away from nuclei.

If you have modelled a reasonable TS guess, you can probablly use that geometry and search for a TS with, say, M062X or dispersion corrected B3LYP and TPSSh to name a few. If you dont find it, then you can start thinking about other TS possibilities, which is already interesting - maybe the reaction doesnt go the way you thought. If you do find it, then you should attempt to characterize it vibrationally through frequencies calculation and, if possible, a complete IRC.

dear Senal,

Semi-empirical methods are usually only recommended for big systems with a large number of atoms, since it has a lower computational cost, but for small systems (less than 50 atoms), ab initio is usually the best choice.