How to determine if item-to-item similarity is high (relatively), using cosine similarity

I’m trying to explain part of my recommendations based on item similarity.
I have a trained model based on Item-Item cosine similarity that performs well. Given two items, ItemA and ItemB, I can compute their cosine similarity. By definition, a similarity of 1 is considered extremely similar, while 0 is the opposite.

Given that my data contains many items, how can I determine how similar ItemA and ItemB are, compared to the similarity between all other items?