We outline a new implementation of a minimal energy conical intersection (MECI) optimization algorithm within the context of semiempirical methods. Computationally, this semiempirical conical intersection optimization method is much less demanding than ab initio CASSCF and MRCI techniques. We apply the method to several molecules and compare the geometries and energies of the resulting MECIs with ab initio CASSCF methods. The locations of the semiempirical MECIs agree very well with the ab initio predictions, but the energetics generally do not. This suggests that the semiempirical conical intersection optimization method may be useful in finding initial guess geometries for ab initio MECI searches and/or in identifying families of MECIs that may be relevant in photochemical dynamics. Indeed, in the present work, we have located many new MECIs for some of the studied molecules that were then verified and refined with ab initio electronic structure theory. The good agreement of MECIs locations further suggests that in many cases, reparametrization of semiempirical methods to reproduce both energetics and locations of MECIs may be successful.