I'm a professor in ML/AI at a Canadian school that's probably roughly in the same caliber for AI as the #15-25 schools on the US News list. (I got my PhD at the #1 school on that list, but at a time when admissions in this area was far easier – I am quite confident I would not get in today with the application I sent in 2011. Even then, I was also rejected from some schools ranked far lower on that list than the one I went to.)
I don't have the numbers in front of me, but admissions rates in AI/ML are actually far worse than for CS grad programs in general: in my department, more than half of the applicants currently say they want to do broadly-AI topics, while even judging very generously as to who counts, under 1/5 of the professors are in that area.
It sounds like you have a first-author AAAI paper – congrats! This is definitely a positive signal that makes me pay more attention to an application. Depending on what area of AI you're talking about, though: "core ML" people like me tend to value AAAI papers significantly less than those from venues like NeurIPS, ICML, or maybe ICLR/AISTATS/UAI, all of which have higher numerical acceptance rates but whose average ML paper tends to be (in my opinion) of higher quality than the average AAAI ML paper. (This is not the case for other areas of AI, though, where AAAI is the premiere venue.)
But: you say you've already done a master's. Having a paper out of undergrad is definitely a distinction from many applicants who do get in (though not everyone agrees), but out of a master's, it's far more typical. (I would expect that most, though certainly not all, admits to top graduate programs who already have a master's will also already have a paper or two.) It's for sure a good thing, but not something that sets you super apart from the competition.
So, this leaves all the other factors that everyone is talking about. The middling GPA from not-well-known institutions certainly hurts. Maybe your reference letters weren't as positive as they could be, particularly if they're from people and/or a country that US schools don't know so well. Maybe the topic of the paper and/or what you wrote about in your statement is different enough from potential supervisors' interests that they have a hard time seeing themselves working with you.
Candidates who are "pretty good but not super-outstanding" can tend to have a hard time in admissions, since the strongest-on-paper applicants will tend to get a lot of offers, but advisors are wary of over-offering since then if too many students come they'll have to find a way to pay for them all.
(As you've mentioned elsewhere, it's highly unlikely you're going to get any specific feedback on that from anywhere.)
In any case: it is what it is. You can either go to the school you're not so thrilled about, or decide to do something else with the few years / rest of your life. The quality of the work you do in grad school is more important than where you do it, and though this will make things more challenging versus going to a top place, this is definitely not necessarily the end of your academic career or anything – plenty of excellent researchers went to lesser-known schools.