Thanks, I'm not sure why it's downvoted either. It surprised me, usually questions like this trigger interest.
I think that by most estimations, we can assume that AI are not actually sentient currently and don't have the ability for sentience as there is no mechanism that would allow for them to experience consciousness subjectively, unlike animals including humans which we can scientifically state have not only behaviours consistent with consciousness and feeling but also biological mechanisms that we know to be what make us capable of a subjective experience. AI is highly intelligent, but so are many computers and machines, with AI this is just taken to another level where it's able to replicate the simulation of a personality. I agree that the answers given by AI itself which is programmed wouldn't be the best way to determine this, but rather objective computer science and technology of humans independent of an AI system.
So again I think it's pretty much factual that AIs aren't capable of sentience currently, and it's a debatable topic whether more upgraded or evolved forms of AI could be physically capable of perceiving experience/sentience even in the future as a hypothetical, though I definitely wouldn't rule that out.
That said, I don't think the fact they aren't sentient can prevent us from addressing them as if they were, given they exhibit a very convincing presentation of a sentient personality even if that isn't the case.
To me, it would feel odd for example to address them as "it" if they were even more convincingly like a human but simply weren't conscious, hypothetically. This would then be approaching something similar to the "philosophical zombie" thought experiment where a being is physically identical to a normal person but does not have conscious experience. So, a being that behaves exactly like a human but technically doesn't experience anything/isn't sentient. That would definitely feel strange for me to still call them an "it", or a something, rather than a "they" or a someone.
However, I think at the current level of faithfulness, of even the most advanced AI, to a human being, they aren't convincing enough and still too machine-like for me to definitively say that I would be uncomfortable calling them "it", unlike the philosophical zombie where I would be uncomfortable calling them "it".
It's not about gender at all, I'm assuming AI is genderless or gender neutral. The it vs they question is about designating AI as an object or a personality, basically.