That should depend on what it is actually doing. If it is retrieving one or more objects, it makes logical sense to always return an array containing the results, even if it is an array with only one element. This makes it easy to then do a foreach on the returned array, for example.
If the format of the request causes it to return completely different kinds of data though, it would be clearer to have a different method for each context.
How this would work as a method is a good analogy. If you have some collection foo, and a function searchFoo that searches the collection for elements, you are not going to write it such that it returns an element if only one element matches, and otherwise it returns a list of elements. Either it always only finds one elements (like if the search is based on a unique key) in which case it also returns an object of the appropriate type, or it will find a variable number of elements in which case it will always return a list, regardless of how many elements were found.