Assume everything relevant is an integer.
$\gcd(even,even) = even$[1]
$\gcd(X, odd) = odd$.[2]
$X\cdot even = even$[3]
$odd \cdot odd = odd$[4]
$odd \pm even = odd$ but $same \pm same = even$.[5]
Those should get you the answer.
$\gcd(even + even, even*even) - \gcd(even,even) = \gcd(even,even)-\gcd(even,even) = even - even = even$.
$\gcd(even+odd, even*odd) -\gcd(even, odd) = \gcd(odd, even)-\gcd(even,odd)=odd-odd = even$.
$\gcd(odd+odd, odd*odd) - \gcd(odd,odd) = \gcd(even, odd)-\gcd(odd,odd)=odd -odd = even$.
So, yes, $\gcd(x+y, xy) -\gcd(x,y)$ is always even.
........
[1] through [5] are obvious, yes?
[1]. Even numbers are divisible by $2$ so their $\gcd$ will be divisible by $2$.
[2]. Odd numbers do not have $2$ as a prime factor so no common factor with an odd number will have $2$ as a prime factor.
[3]. Every multiple of a multiple of $2$ is a multiple of $2$.
[4]. a) euclids lemma says if $2|x,y$ then $2|x$ or $2|y$ so if $x$ and $y$ are both odd $2|xy$ is impossible. (Are we could do the elementary way: $(2k+1)(2j+1) = 2(2kj +k + j) +1$.)
[5]. Hmmm... it'd be fun to come up with the single slimmest line to argue this. Not sure what the slickest argument is, but some not slick ones are obvious.
$X + even = parity\ of \ X$ because $2|even$ so $2|X+even \iff 2|X$. and $X + odd = opposite \ parity \ of \ X$ because $2\not \mid odd$ so $odd \equiv \pm 1 \pmod 2$ and $X+odd \equiv X\pm 1 \equiv \begin{cases}0+1=1\\1-1=0 \end{cases}$.
But that's the exact opposite of slick.
Guess we should go with the inelegant: $same \pm same = (2j+\begin{cases}0\\1\end{cases}) \pm (2k\mp\begin{cases}0\\1\end{cases})= 2(j\pm k)=even$ whereas $odd \pm even = (2k+1) \pm 2j = 2(k\pm j) + 1 = odd$.