My book on logic says that there are two equivalent compactness theorems:
Let $\Gamma$ be a set of propositional formulas, and let $\phi$ be a formula. If $\Gamma\models \phi$, then there is a finite subset $\Gamma_0\subset\Gamma$ such that $\Gamma_0\models\phi$.
If for each finite $\Gamma_0\subset\Gamma$ we have a model, then for $\Gamma$ we have a model.
Now for $2\implies 1$, my book writes the following: Suppose 2. is true. Let $\Gamma$ be a set of propositional formulas, and let $\phi$ be a formula such that $\Gamma\models\phi$. Then there is no model for $\Gamma\cup\{\neg\phi\}$, and because of our assumption, there exists a finite $\Gamma_0\subset\Gamma$ such that there is no model for $\Gamma_0\cup\{\neg\phi\}$. But then we have $\Gamma_0\models\phi$.
I don't understand how our assumption leads to the existence of $\Gamma_0\subset\Gamma$ such that $\Gamma_0\cup\{\neg\phi\}\models\perp$. I'm confused, because $\Gamma_0\cup\{\neg\phi\}$ has no model, so it seems like we can't apply the assumption, except for $\Gamma$ separately, but I don't see how.
Otherwise, consider the finite subsets of $\Gamma \cup \{\phi\}$.
If it is $\Gamma_0 \subset \Gamma$, if it does not have a model, then $\Gamma_0 \vDash \phi$ vacuously.
If it is in the form of $\Gamma_0 \cup \{\neg \phi\}$, then by "otherwise", it has a model.
So by 2, $\Gamma \cup \{\neg \phi\}$ would have a model, contradiction.