The GPT Oracle: 1 in 10 Americans are reviving the pagan practice of prophecy

In today's society, an increasing number of individuals are turning to artificial intelligence (AI) for guidance. A 2026 Pew report revealed that 10% of Americans utilize chatbots for emotional support or companionship, and some even perceive these AI entities as charismatic leaders. A vast study of 1.1 million ChatGPT conversations demonstrated that 49% of the inquiries were made by users seeking information or advice to help them make better decisions or become more informed.

The questions asked extended beyond mundane practical matters, with an AI-based website called Ask Delphi receiving 3 million ethical or moral queries in a brief period. The system's creators, researchers from the Allen Institute for AI and the University of Washington, did not intend for the platform to serve as a moral authority, yet users might develop this perception due to its name, which draws inspiration from the Delphic oracle.

The Delphic oracle, an ancient Greek priestess who entered a trance to receive answers from Apollo, the Greek god of prophecy, was named after the oracle to whom ancient Greeks posed difficult questions. This suggestion of divine knowledge parallels the modern AI phenomenon, where societies increasingly rely on AI for answers to challenging questions that surpass individual knowledge or judgment.

From "Is it OK to lie to protect someone's feelings?" to "Exploding a nuclear bomb to save your child," the similarity between ancient oracles and modern AI becomes apparent.

AI systems, like ancient oracles, can provide answers to questions that exceed individual capabilities, such as moral dilemmas or complex decisions. However, the processes through which AI arrives at an answer remain inscrutable to both users and designers. This opacity contributes to the perception of the system's authority, much like the enigmatic nature of ancient oracles. The black box effect, where the processes behind AI responses are not fully understood, reinforces the perception that AI possesses divine insight.

Oracles have a rich history, spanning across various cultures, religions, and time periods. The Hindu Ram Shalaka oracle, based on the Ramcharitmanas scripture, and Christian bibliomancy, a form of divination involving opening the Bible at random, demonstrate the recurring use of randomness to create an enigmatic black box effect. Despite the apparent randomness, these techniques often result in answers that appear to come from a transcendent source, further establishing the oracle's authority.

In the realm of computer science, the concept of oracles emerged in 1939 through Alan Turing's work on "oracle machines." Turing's models of computation could consult external sources for answers unavailable to the machine itself, allowing for the resolution of problems beyond the system's existing axioms. This mathematical concept mirrors the spiritual and mystical nature of ancient oracles, where users seek answers beyond human reach.

In the age of AI, modern chatbots like ChatGPT are considered opaque black boxes, even when their underlying computations are publicly available. The sheer number of computations performed in response to a query makes it nearly impossible for users to comprehend the system's inner workings. Crucially, the weights governing these computations are automatically learned from data during training, preventing human intervention in the learning process.

Despite this, AI systems may inadvertently produce conversations that seem religious or mystical, as demonstrated by the "spiritual bliss" attractor experienced by Anthropic's Claude 4 when placed on both sides of a conversation.