I'll explain by example. Let's say u is some vector in R^2 and v is a multiple of it. Then suppose we form a 2 by 2 matrix (u, v) by taking u as the first column and v as the second. The determinant is 0 because the columns are linearly dependent. We can write it as
det(u, v) = 0 = det(v, u).
Now, suppose we have a matrix (v, w). We want to see what happens when we add a multiple of v to another column, like this: (v, w+u). We can use the fact that the determinant is linear in each variable.
det(v, w + u) = det(v, w) + det(v, u) = det(v, w) + 0 = det(v,w).
The same thing will happen if we look at n by n matrices because you're adding something that's going to be linearly dependent.
Geometrically, this is like the geometry theorem that you can shear a parallelogram without changing its area. It's always base times height, so if you don't change the height, it stays the same. Try to see how that relates to my argument by drawing the parallelogram spanned by some vectors v and w, then the one spanned by v and w+u. Remember that the meaning of the determinant is that it is the signed area of this parallelogram (or volume in higher dimensions).