As someone whose entire magical life has taken root and blossomed in online spaces, I was happy to get the opportunity to review Davezilla's Magical AI Grimoire: a Book of Shadows for Contemporary Chaos. Alas. Sometimes, a book tries very hard to make me not like it. More to the point, the people that surround the book do it: the author, the publisher, maybe even the writer of the forward or anyone else that involves themself in the marketing process. When the publisher adds a preface which says in the opening paragraph that this book "dares to look at the role of AI in magical practice" I have to admit my own biases, not towards artificial intelligence but, instead, towards people who are concerned with how edgy they must be. When the writer of the foreward believes that the "only true creative force" is chaos magic, my eclectic background feels like its fur got rubbed against the grain. That said, one of the things a person tries to do if they've spent decades in online spaces is to filter out the noise of the messengers and see if there's good information that's simply being badly delivered. After all, just because the author says things like "Google it" unironically it doesn't automatically mean that they are a dismissive jerk, they could just be embracing the online idiom to appeal to the book's target audience.
That was not a flower-bestrewn introduction, so let me go on by saying that Davezilla is a good person to have written this book. He is a passionate advocate for the technology, understands what it is and is not (for example, he very early on clarifies that almost all of what we have access as of the publication in 2025 are more accurately described as "Large Language Models" or "LLMs" and not truly "Artificial Intelligence") and has a background that has encompassed numerous aspects of metaphysical genres; that's a queer turn of phrase but he spends a fair amount of the Introduction section breaking down how many different names there are for precisely-defined corners of this world so a catch-all term is tricky to use respectfully. Also, you can't come away from reading this book without thinking that he's put a lot of thought into writing prompts that facilitate a practitioner who wants to make use of these tools. I can't claim to be Davezilla's technological equal but to the best of my understanding their advice in the use of prompts is found. From formatting tips (and the variant requirements bases on which version of currently available tech you have access to), to thoughtful descriptions of what you want the LLM output to achieve, you can follow their work without hesitation. If, for example, you want a healing spell tailored to your non-traditional Wiccan practice, you can turn to page 99 and, sure as anything, find a prompt crafted to generate what you're looking for. There are numerous examples covering a wide spectrum of both traditions and types of magic. Seeing the differences in the prompt he writes to get a spell to aid job hunting, but one for a "Traditional Witch" and another for a "Chaos Magician" really helps to educate the willing learner.
Fundamentally, you should consider this review as an endorsement of the book. That being said, I'd also recommend putting the book on your shelf and letting the future catch up to it for awhile. The author (and the publisher, and the writer of the introduction) take it as settled that these LLM tools, and the AI tools that will inevitably follow, are ethical and we'll all come to accept that eventually. What never gets talked about enough, in my opinion, is that when (for example) the LLM provides you with a Discordian protection ritual, it learned/synthesized that information from sources that it may or may not have had any right to have accessed. Meta (the parent company of Facebook) is revealed in court documents as having pirated ~82 terabytes of books to feed to its LLM efforts; the cynic in my finds it hard to believe either that a) that is their only offense or that b) they are the only one of the competing corporations to behave in this fashion. The typical capitalist attitude often seems to be "well this might not be ethical but how else will my genius idea make the shareholders billions of dollars? You want me to pay for all this intellectual property or something?" So many closed practices have had their secrets broached, but it's one thing for you to decide which information on Orishas you can ethically make use of and which you should not; it's an entirely different problem to know which of those materials was sieved through your Large Language Model of choice to provide you with the spell you prompted it to create. Unless you take the author's glib attitude towards these things I would propose that this is far from settled science, and so I can both encourage you to read Magical AI Grimoire as an academic exercise while simultaneously try to dissuade you from putting much of the work into practice.
~ review by Wanderer
Author: Davezilla
Weiser Books, 2025
288 pg., $22.95
