What you're describing is called a running key cipher. A running key cipher turns into to a one-time pad (and thus becomes unbreakable) if the key is truly random, kept secret, and not reused. Quoting from my article at the end of the first section: "Running-key ciphers are constructed similarly to one-time pads, but instead of using a truly random key, they usually use text (from say a newspaper) as the key. Although these ciphers can still be difficult to break, an adversary like Eve can use patterns in natural language, or intelligence about probable messages or texts to decrease the difficulty of guessing the right key/plaintext."

Eve could guess likely plaintexts/keys and subtract those from the ciphertext until she finds something intelligible. It would be quite probable that this snippet of intelligible language is from either the plaintext or key. At this point it becomes fairly trivial to extend either one using what she knows about Alice or Bob, possible texts they could be using as a key, or just features of natural language.

Another problem arises when we consider the computational capacity of an adversary in today's security environment. Nation-states' signals intelligence apparatuses (and some large corporations) have access to more than enough computational resources to trawl through most available literature to find the key. Quoting from the wikipedia page on running key ciphers: "Suppose that there are 100 million key texts that might plausibly be used, and that on average each has 11 thousand possible starting positions. To an opponent with a massive collection of possible key texts, this leaves possible a brute force search of the order of 2⁴⁰, which by computer cryptography standards is a relatively easy target."

So unless the stakes are low, relying on a running key cipher is not advisable. Thanks for commenting, and hopefully this helps!