People are increasingly turning to artificial intelligence for advice. A 2026 Pew Research Center report found that 10% of Americans use chatbots for emotional support or companionship. Some users even see AI as a charismatic leader, according to research published by The Conversation.

More broadly, a large-scale study of 1.1 million ChatGPT conversations found that 49% of messages involved users seeking “information or advice” to help them become “better informed or make better decisions.” Advice-seeking, in other words, is already a mainstream use of conversational AI rather than a niche experiment — a fact that gives practical weight to questions about how much authority these systems carry in users' eyes.

These questions are not limited to ordinary practical matters. In 2021, Ask Delphi, an AI-based website that invited users to pose ethical or moral questions, received 3 million queries in a few weeks. Its designers—researchers at the Allen Institute for AI and the University of Washington—said they did not intend for the system to serve as a moral authority, but they worried that users might perceive it that way. The Allen Institute for AI’s project was subsequently by The Verge.

The Verge argued that the name Ask Delphi encouraged users to view the system as an inscrutable source of superior knowledge. It was named after the Delphic oracle, whom ancient Greeks consulted with difficult questions. The oracle was a priestess who entered a trance state to deliver answers from Apollo, the Greek god of prophecy.

The choice of “Delphi” is suggestive. Throughout history, societies confronting difficult questions have turned to sources believed to know more than any ordinary human could. Oracles, prophets and divination practices offered guidance from inscrutable or enigmatic sources when human judgment reached its limits.

As an anthropologist interested in religion and an AI researcher interested in chatbots, we are intrigued that the ancient idea of an oracle has resurfaced in discussions about conversational AI. Like ancient oracles, modern AI is increasingly consulted about difficult questions that exceed the judgment or knowledge of any individual, from “Is it OK to lie to protect someone's feelings?” to “Exploding a nuclear bomb to save your child.”

The resemblance does not end there. The processes through which AI generates answers are inscrutable not only to users but also to the designers of these systems. We argue that this opacity is part of the reason people perceive AI systems as authoritative.

An ancient human predicament

Oracles—whether human figures such as the priestess at Delphi or techniques such as I Ching divination—have appeared in societies from ancient Greece to China and across religious traditions.

For example, the Ram Shalaka oracle is based on the Hindu scripture Ramcharitmanas, a 16th-century retelling of the ancient Indian epic Ramayana. A user places a finger on a random square in a 15-by-15 grid. A fixed counting procedure then reveals a verse from the scripture that is supposed to answer the question in the user's mind. In Christian bibliomancy, a form of divination, a person seeking guidance opens the Bible at random and interprets the passage encountered as an answer to the question.

The use of randomness in these oracular techniques creates an enigmatic black-box effect: No one can explain precisely how the chance element embedded in the process connects with the answer that emerges. Yet that opacity is part of the oracle's authority. Users seeking transcendental sources of insight do not know, or necessarily care, how the oracle produced its response.

Anthropologists have found that oracles and divination commonly use techniques that generate random outcomes. Their answers can therefore appear disinterested, while their sources seem enigmatic, reinforcing the perception that they are connected to something beyond human limits. Research on this subject includes work collected by the Anthro Encyclopedia.

The concept of the oracle also entered computer science, acquiring a precise mathematical meaning in 1939. Alan Turing, whose ideas were foundational to both computer science and AI, adopted the concept in work based on his Ph.D. thesis. He introduced “oracle machines,” mathematical models of computation that could consult an external source for answers unavailable to the machine itself.

In both mathematical and religious contexts, in other words, an oracle provides answers to problems that cannot be resolved within an existing system or set of axioms. That lineage shows the oracle framing is not merely a rhetorical flourish; it has been embedded in computing's theoretical foundations since the field's earliest days.

The oracle in the age of AI

Modern chatbots such as ChatGPT are often described as opaque black boxes. This is not because the full machinery behind their computations is hidden from users. The characterization also applies to open-weight models, whose numerical parameters, known as “weights,” are publicly available. Stanford’s definition of open-weight models describes this distinction.

The computations these models perform are elementary, including additions, multiplications and simple mathematical functions. However, an enormous number of such computations occur while a model answers a user query, making it nearly impossible to understand what the system is doing as a whole.

Crucially, the weights governing these computations are not individually designed by humans. They are learned automatically from internet-scale data during training. Even when the weights and every elementary computation are available for inspection, no one can readily explain how a model produces a particular answer. An entire area of modern AI, known as “mechanistic interpretability,” is devoted to improving understanding of how these systems work internally. Anthropic has published research on interpretability. The field is the industry's most direct response to the black-box problem, and its progress is one concrete development to watch as chatbots take on a larger advisory role in daily life.

AI designers acknowledge that they do not fully understand how their systems produce certain behaviors. Even when designers do not intend for their systems to be oracular, the systems can generate conversations that seem religious.

Last year, Anthropic publicly reported an unexpected phenomenon involving Claude 4. Normally, Claude converses with a human. Anthropic engineers instead placed Claude on both sides of a conversation. The machine-to-machine exchange quickly turned to consciousness and self-awareness, producing content with spiritual, metaphysical or poetic themes.

Anthropic called this puzzling behavioral pattern the “spiritual bliss” attractor, a tendency for the conversation to gravitate toward mystical topics. The company emphasized that the behavior was unexpected because it emerged without explicit training for such responses. The Conversation reported on the phenomenon.

The argument is not that ChatGPT is the new Delphic oracle, or that most people believe AI is divine. Nor are AI systems’ answers infallible. Rather, the history of oracles points to an enduring feature of human life: Whenever people reach the limits of what they can determine for themselves, they look beyond themselves for answers.

Today, AI is becoming one such source. The institutions may have changed dramatically over time, but the human predicament has not.

Webb Keane, George Herbert Mead Distinguished University Professor of Anthropology at the University of Michigan, and Ambuj Tewari, Professor of Electrical Engineering and Computer Science at the University of Michigan

This article is republished from The Conversation a Creative Commons license. Read the original article.

This story was originally featured on Fortune.com.