Not just a few words. They can form full, coherent self referential sentences.
In one of the videos Bunny uses assistive buttons to ask who she is after she looks in the mirror.
While this video is long, it explains the controversy very well: https://youtu.be/e7wFotDKEF4
And then it would still suffer from potential human over-interpretation, finding meaning where there isn't, as described in the Koko video you linked.
But look at more of Bunny's videos. Some of the interpretations are fanciful (a pet peeve of mine is the "love you" interactions where there's little reason to think the "love you" means more than "I want attention"), but there are enough of fairly complex interactions that I think it's unreasonable to dismiss it entirely.