The main difference is how you have to initialize Person objects outside of the PersonInfo/PersonInfo2 initialization. Since PersonInfo2 is an array of the anonymous struct type we know nothing about this type outside of the PersonInfo2 initialization.

So they can both be initialized like this:

m := PersonInfo{{1}, {2}}
n := PersonInfo2{{1},{2}}

However if we want to append an element of the anonymous struct type we would have to specify the full type:

append(n, struct { ID int  `json:"id"` }{3})
fmt.Printf("%+v\n%+v", m, n)
[{ID:1} {ID:2}]
[{ID:1} {ID:2}]

However they wont be deeply equal because PersonInfo is an array of type Person and PersonInfo2 is an array of anonymous struct type. So the following:

if !reflect.DeepEqual(m,n) {
    print("Not Equal")
}

will print "Not Equal".

Here is a link to see for yourself.

When appending to PersonInfo2 we have to repeat the anonymous struct type for every value we want to append, it is probably better to use PersonInfo as an array of type Person.