I'd sort of say its two fold:

1. the object in question would need to behave in a conscious way 2. I'd have to look in the guts of the object and see evidence of some kind that the processes I was observing were inherent in the design of the thing.

To clarify a bit about the second point: I can talk to a phone with a person on the other end of it and it can demonstrate some properties of consciousness. If I didn't know what phones were I'd be tempted to identify this behavior with consciousness if not for the second criteria. In the case of a phone, if we take it apart and see how it works we can see that it is fit to transmit a voice from somewhere else, but not to generate the complex responses I was observing in the conversation.

Obviously the case of a language model, for instance, is significantly more complicated, since it does have some of the behaviors of a consciousness and some of the structures, but my hunch is that it doesn't really have enough of either to count. Definitely not discounting the possibility that we'll have artificial consciousnesses some day, I'm just skeptical language models are it.