Assume that $gcd(a,b)=d$. This leads to conclusion that $a=k \cdot d$ and $b = n \cdot d$
Now let $gcd(k,n) = m$, then we can write: $k = s \cdot m$ and $n = t \cdot m$. Substitute back we have: $a = s \cdot m \cdot d$ and $b = t \cdot m \cdot d$. Both $a$ and $b$ have $md$ asa factor, but we know that $gcd(a,b) = d$, so that implies $m=1$
Q.E.D.
Also another way is to use the following property of the function Greatest Common Divisor:
$$gcd(ma,mb) = m \cdot gcd(a,b) \forall m \in \mathbb{Z}; m \neq 0$$
Let $a = kd$ and $b=nd$. Then from assumtion we have:
$$gcd(a,b) = d$$
$$gcd(kd,nd) = d$$
$$gcd(k,n) \cdot d = d$$
$$gcd(k,n) = 1$$
Q.E.D.
To make it clear the Greatest Common Divisor is as I would like to call it "semi-multiplicative" function, i.e. it's multiplicative, but only under specific condition. For example, this will hold:
$$gcd(ma,b) = gcd(m,b) \cdot gcd(a,b)$$
if $gcd(m,a) = 1$
So your calculation is wrong because it true that:
$$gcd(\frac ak, \frac bn) = d$$
But you multiply both sides with $nk$, that will change the value of The GCD, unless $(n,k) = (a,n) = (b,k) = 1$, but we can't make that restriction.
Also that division will never lead to an answe that $gcd(n,k)=1$. Because we can make the following factorization if $(a,n) = (b,k) = 1$
$$gcd(a \cdot n, b \cdot k) = (a,b) \cdot (a,k) \cdot (b,n) \cdot (n,k)$$
But obviouly we can't make that restriction. For counterxample take $a=12$ and $b=18$.This will provide $d=6$, $k=2$, $n=3$.
So in your example you'll have:
$$(a\cdot n, b \cdot k) = (36,36) = 36 \neq d = 6$$