Our recommend algorithm uses artificial intelligence to train the models. The frequently bought together model will recommend items that were bought together and some that might go well together. For example, if item A was was bought with item B, and item B was bought with item C, the model might learn to recommend item C to A or item A to C.
Furthermore, if you are using an "add to cart" conversion event, it is possible that some noise will appear if a user adds an item to cart but does not complete the purchase. These events will still be considered in the recommend model.