By definition, one can always know their own naturally-produced neural net is self aware (cogito, ergo sum). The open question for solipsists is whether, should someone report to you their neural net is also self-aware, you should believe them.

but how can we be sure that the cogito part is actually happening? this immediately (d)evolves into a purely ontological discussion

Personal experience. We perceive a continuity of experiences, memory, and the difference between that and slumber or becoming insensate, and we call the experience "consciousness" and our knowledge of that "self-awareness." It is definitional; the word descends from the experience, not the other way around.

Perhaps we need to synchronize our definitions of "experience", "perception", "personal", and "knowledge", because your very words are almost exactly what I would say in counter to your point.