Lie algebras are usually classified modulo isomorphisms of the underlying vector space, i.e. invertible changes of basis, which are $GL(N)$ maps of the generators $X'^i = M^i{}_j \, X^j$. For example, there are only two classes of 2D Lie algebras: the Abelian one $[X,Y]=0$ and the non-Abelian one $[X,Y]=Y$. A generic 2D Lie algebra would be $[X,Y]=a \, X+b \, Y$, but unless $a=b=0$, there always is a $GL(2)$ transformation like above, such that $[X',Y']=Y'$. Notice that such a transformation in general is not an algebra homomorphism: it does not preserve the Lie brackets (i.e. it does not leave the structure constants unchanged).
Lie algebra automorphisms are homomorphisms from the algebra to itself, that is, linear maps $X^i \to \varphi (X^i)$ such that $[\varphi(X^i),\varphi(X^j)]=\varphi([X^i,X^j])$. They are a subset of the changes of basis that preserve the Lie brackets, or, equivalently, that preserve the structure constants. In the example above, an automorphism of the Lie algebra $[X,Y]=Y$ is the map $X'= X$, $Y' = \alpha \, Y$ with $\alpha \neq 0$. This map is such that $[X',Y']=Y'$. Clearly not all $GL(N)$ changes of basis are automorphisms, and they form a nontrivial subgroup of $GL(N)$.
Now, when Lie bialgebras are studied, their classification is usually done up to automorphisms of the underlying Lie algebra. That is, if I want to classify all Lie bialgebras on the Lie algebra $\mathfrak g$, two of them are considered equivalent if there is an automorphism of $\mathfrak g$ that sends the cobracket of one into the other. My question is, why not consider equivalent two Lie bialgebras on the same vector space if they are related by a generic $GL(N)$ change of basis? Why limit it to automorphisms? The way that $GL(N)$ changes of basis act on the structure constants of the algebra and of the coalgebra, they all preserve the validity of the Jacobi, co-Jacobi and compatibility rules, whether they are automorphisms or not. In other words, any change of basis sends Lie bialgebras into Lie bialgebras, so it seems natural to me to consider equivalent two Lie bialgebras that are related by any $GL(N)$ transformation. What am I missing?