The AI tagging is the best I have seen, which is to say impressive but ultimately not incredibly useful.
I was impressed with some of the classifications it made, e.g. picking out parrots from other birds, but as ever with AI there were plenty of classifications that were not made. So most of the shots I checked that had “parrot” did have a parrot in them, but not all parrot shots were classified thus.
The real test is Excire’s “find similar” function. Pick a photo that impresses you with its classification, then click the magnifying glass. I did this once and actually laughed out loud.
As I stated in their forum, AI won’t be of practical use until it’s something like 90% accurate and that is many years off yet. For my own purposes, I do not expect to still be alive when AI can do automatically what I do with keywords today, even at 50% accuracy.
Sadly, the AI seems to be their focus and I find this regrettable. AI features can materially add to the appeal of a product, but pinning everything to the AI banner is, in my view, a quick road to oblivion.
An example of good AI use is Pixelmator Pro’s AI Super Resolution which does an incredible job of upscaling images, but that is one feature in a very well rounded application. Likewise Pixelmator Photo’s ML adjustments are a useful addition but you also have full manual control of everything ML can touch.