I think people are generally aware that the fundamental issue is experience of suffering, as you say. It's just that since that can't be measured directly, people talk about intelligence and pain reaction since that's what is studied and observed in order to infer about experience of suffering.

But for synthetic intelligences, I think we need to be extra careful because the suffering can be there without any means to notify humans, if those means were never programmed in. If an AI has self awareness and self preservation goals comparable to humans, but no comparable e.g. English language subroutines, being shut off at random times would probably cause it to suffer psychological distress in silence.

This is all to say that for the sake of both AI and animals, we need to learn how to directly detect brain activity corresponding to pain, in different types of brains.