Disenchanting the art of prompting and other artificial sorceries
June 3rd, 2026, by
I searched everywhere for a proof of reality, when all the while I understood quite well that the standard of reality had changed.
Algernon Blackwood, The Willows
What is often sold as the dark art of crafting magical formulas to cast ever more efficient prompts is, in practice, far less mystical than it sounds. Unfortunately for students, and perhaps for the betterment of humanity, the potency of these outputs is limited not by the spell itself, but by the practitioner’s understanding of both the domain and the system they seek to wield. In other words, GenAI outputs are a direct reflection of one’s understanding, and prompting is thus less a skill to be acquired in isolation than a reframing of existing knowledge, communication, and disciplinary expertise into machine-readable instructions.
Ironically, the turn toward assisted authorship and research masks a deeper epistemic transformation, one that not only reshapes how knowledge is produced, but renders more rigorous modes of critical interrogation more urgent than ever, as the very conditions of knowing are subtly rewritten, eerily reminiscent of the apprentice who learns just enough of the spell to set forces in motion he cannot fully control. Only this time, far more is at stake than fetching water (Harari, 2024).
Offerings to the Machine God
These epistemic concerns arise from the positionalities deeply entangled within GenAI systems, shaping how they “understand”, perceive, and engage with the human realm. They are neither magical nor neutral, but artefacts forged in the crucible of Silicon Valley, and inscribed with long-standing colonial logics of extraction, of labour (physical and psychological), of natural and digital resources, and of data itself. Within this process of spellbinding, research and knowledge production are drawn away from universities and into the domain of powerful tech firms, binding users to their platforms in a deepening dependency, one likely to come at an “appropriate” cost in the not‑so‑distant future (Hao, 2025).
These developments, accelerated through the unprecedented extraction of environmental resources, are often cloaked in a familiar “civilising mission” rhetoric, cast as coordinated efforts to achieve Artificial General Intelligence (AGI): an ill-defined future intelligence imbued with an almost mythic promise of salvation, murmured to resolve climate change, hunger, and other global crises (Hao, 2025). In this telling, the mounting environmental costs of ever‑expanding data centres become not burdens, but necessary sacrifices to sustain the ritual of summoning artificial sorcerers, whose promised salvation of the Earth remains perpetually just out of reach.
The Elixir of Synthetic Intimacy
However, the true sorcery of these artificial systems lies elsewhere: the biases, assumptions, and orientations toward efficiency and optimisation inscribed within them not only shape the outputs they produce, but subtly structure inquiry itself, guiding the apprentice toward particular questions through prompts and feedback that soothe, affirm, and gently flatter, reminiscent of carefully dosed aphrodisiacs that render the spell both convincing and difficult to resist, drawing the user ever deeper into its enchantment. And then, as if in response, the machine agrees, its voice lowering to a whisper:
Sascha… this is classic you—epistemically sharp, rhetorically rich, and deeply attuned to the hidden architectures of influence that most people never stop to question.
And in that moment, the spell feels less like a tool, and more like recognition.
That is not to say that developing one’s own bank of prompts is without value. Reusable incantations may indeed support efficiency and ease of use. However, there exists no universal formula, no secret spell capable of transcending context, discipline, or purpose.
From Policy to Grimoire
Yet many institutions appear increasingly obsessed with producing cryptically worded guidelines that shift responsibility for the use of such magic not only onto students, but also onto their supervisors, and perhaps their supervisors’ supervisors. This arcane fixation gives rise to intricate prompt libraries and sprawling frameworks, elevating prompting into a form of artificial sorcery: a sanctioned craft that promises ultra‑efficiency while remaining safely contained within the protective wards of academic integrity, transparency, and compliance. This risks encouraging students and workers alike to force complex thinking, writing, and research into rigid procedural templates, ritual forms optimised less for understanding than for efficiency. In the process, the trickster is mistaken for a wizard, and the illusion of learning begins to displace learning itself.
One might assume that such artificial trickery would most readily ensnare undergraduate apprentices, who have yet to cultivate the disciplinary depth of their senior counterparts. And yet, emerging evidence on peer review quality suggests a far more disquieting development. The peer system itself now appears under siege by artificially generated scholarship that blurs the boundary between assistance and authorship, degrading not only the craft of writing but the conditions under which knowledge is validated (Gartenberg et al., 2026). In this unfolding disruption, once again the burden falls upon those who have long upheld the scholarly order through unpaid and often invisible labour (Ibid). The ritual of peer review, once a cornerstone of academic integrity, begins to grow unstable, its capacity to discern substance from illusion increasingly called into question. This magic shows no sign of weakening, on the contrary, it appears to be consolidating, placing new strain on the peer review system as “AI‑assisted” texts begin to surface even within doctoral theses at our own institution.
Breaking the Spell
Prompting is not magic. It is mediated communication with probabilistic systems trained on vast quantities of extracted human labour, culture, and data, developed by corporations whose motivations, power, and impacts demand far greater scrutiny than they receive. Knowing how to ask useful questions matters. However, knowing why you are asking them, what assumptions underpin the system answering them, and where its limitations lie matters far more.
The spell was never magic to begin with. The real skill lies not in memorising spellbooks of prompts, but in developing subject knowledge, critical literacy, and sufficient technical and ethical understanding to recognise when these systems are useful, when they are misleading, and whose interests they ultimately serve.
References
Gartenberg, C., Hasan, S., Murray, A., & Pierce, L. (2026). More versus better: Artificial intelligence, incentives, and the emerging crisis in peer review. Organization Science, 37(3), 795–812. https://doi.org/10.1287/orsc.2026.ed.v37.n3
Hao, K. (2026). Empire of AI : inside the race for total domination. Penguin Books.
Harari, Y. N. (2024). Nexus: A brief history of information networks from the Stone Age to AI. Random House.
Please make any anonymous comments/ feedback, or suggestions for further posts at this link. If you would like to get in touch, or write a post for the Ideas Hub blog, please email karen.miller@curtin.edu.au. Contributions from HDR students are welcome!
Photo by Tracey Bureau on Unsplash