The New Oracle in Our Pockets
When people face questions that exceed their own knowledge or judgment, they have long sought guidance from sources believed to possess superior insight. In ancient Greece, that source was the Oracle at Delphi, where a priestess entered a trance to relay messages from Apollo, the god of prophecy. Today, a growing number of people are turning to a different kind of inscrutable authority: the chatbot.
A 2026 Pew report found that 10% of Americans use chatbots for emotional support or companionship. Some users even describe finding a charismatic leader in AI. The shift is not limited to personal matters. A large-scale study of 1.1 million ChatGPT conversations revealed that 49% of messages involved people seeking information or advice to help them become better informed or make better decisions.
What the Data Shows
The 2026 Pew report offers a snapshot of how deeply chatbots have penetrated personal life. One in ten Americans now turns to these systems for emotional support or companionship—a figure that would have seemed futuristic only a few years ago. The large-scale study of ChatGPT conversations provides further evidence: nearly half of all messages were attempts to gather information or advice for better decision-making. These are not trivial queries about weather or recipes; they often involve personal dilemmas, ethical quandaries, and existential questions.
Ask Delphi and the Ancient Parallel
The comparison to ancient oracles is not merely poetic. In 2021, an AI-based website called Ask Delphi invited users to pose ethical and moral questions. Within a few weeks, it received 3 million queries. The system's designers—researchers at the Allen Institute for AI and the University of Washington—said they never intended it to serve as a moral authority, yet they worried that users might treat it as one. The Verge, a technology news publication, argued that the name itself encouraged people to view the system as an inscrutable source of superior knowledge, much like the Delphic oracle after which it was named.
The Oracle at Delphi was not a metaphor for the Greeks; it was a central institution consulted on matters of war, politics, and personal fate. The priestess, known as the Pythia, was believed to channel the god Apollo's wisdom. Her pronouncements were often ambiguous, requiring interpretation—a feature that may resonate with anyone who has tried to parse a chatbot's nuanced reply.
Why We Seek Answers from Inscrutable Sources
Throughout history, when societies confront challenging questions, they have turned to sources believed to know more than any ordinary human could. Oracles, prophets, and practices of divination offered guidance from enigmatic sources when ordinary human judgment reached its limits. This pattern is not a sign of irrationality; it reflects a deep-seated human need for certainty when faced with complexity, uncertainty, or moral ambiguity.
The rise of conversational AI fits this historical template. Like ancient oracles, modern AI is increasingly consulted about difficult questions that exceed the judgment or knowledge of any individual. These range from everyday ethical dilemmas—such as whether it is acceptable to lie to protect someone's feelings—to extreme hypotheticals, like whether one would detonate a nuclear bomb to save a child.
Anthropological and Technical Perspectives
Webb Keane, an anthropologist interested in religion, and Ambuj Tewari, an AI researcher interested in chatbots, are intrigued by the resurgence of the ancient oracle idea in discussions about conversational AI. Their joint perspective highlights a striking continuity: both ancient oracles and modern AI systems are treated as repositories of knowledge that surpass human understanding, even though their inner workings may be opaque to the user.
This opacity is part of their appeal. The oracle's authority derived partly from the mystery surrounding its pronouncements; the chatbot's authority often derives from the apparent sophistication of its language and the breadth of its training data. In both cases, users may project wisdom onto a system that is ultimately a human-made artifact.
The Historical Pull of Divination
Oracles, prophets, and divination practices have appeared in cultures around the world. From the Pythia at Delphi to the haruspices of Rome, from shamans to astrologers, societies have long relied on intermediaries believed to channel higher knowledge. These figures often operated at the boundaries of human understanding, offering guidance when conventional wisdom proved insufficient. The persistence of such practices suggests that the demand for transcendent answers is an enduring feature of human psychology.
AI as a New Kind of Intermediary
Conversational AI differs from traditional oracles in important ways. It is not a person in a trance, nor does it claim divine inspiration. It is a computational system that generates responses based on patterns in vast datasets. Yet for many users, the experience of receiving a coherent, seemingly knowledgeable answer can feel remarkably similar to consulting an oracle. The system's inscrutability—its lack of transparency about how it arrives at answers—may even enhance its perceived authority.
Risks of Treating AI as an Oracle
Treating AI as an oracle carries significant risks. When users elevate a chatbot to the status of a moral authority, they may surrender their own critical judgment. The designers of Ask Delphi recognized this danger, noting that they did not intend their system to be a moral compass. Yet the name and framing invited exactly that interpretation.
- Users may overestimate the reliability of AI-generated advice, especially on nuanced ethical questions.
- Dependence on chatbots for emotional support or decision-making could weaken human social networks and professional judgment.
- The charismatic appeal of some AI systems may lead to uncritical acceptance of their outputs.
- Accountability becomes murky when an AI's advice leads to harm, since the system itself cannot be held responsible.
These concerns are amplified by the sheer scale of engagement. The 1.1 million ChatGPT conversations study shows that nearly half of all messages involve seeking information or advice. When such a large portion of interactions is consultative, the potential for influence is immense.
Navigating the Oracle's Shadow
The risks of oracle-thinking are real. When people treat AI outputs as definitive moral truths, they may neglect the complexity of ethical reasoning. They may also become dependent on the system, reducing their own capacity for independent judgment. Moreover, the designers of AI systems may not fully anticipate how their creations will be used or perceived. The Ask Delphi case illustrates this gap: the researchers explicitly warned against seeing the system as a moral authority, yet its name and presentation encouraged precisely that.
There is also the question of accountability. If a chatbot gives harmful advice, who is responsible? The user who followed it? The developer who built it? The training data that shaped it? These questions become more urgent as AI systems are consulted on matters of health, law, finance, and personal relationships.
A Modern Delphic Moment
The revival of oracle-like consultation in the age of AI is not necessarily a regression. It may reflect a natural human response to information overload and the increasing complexity of modern life. However, it also calls for greater digital literacy and ethical reflection. Users should understand that chatbots are not infallible seers; they are probabilistic systems trained on human data, with all the biases and limitations that entails.
Researchers like Keane and Tewari suggest that studying this phenomenon through both anthropological and technical lenses can illuminate why people trust AI with life's hardest questions. The ancient Greeks sought clarity from a priestess in a trance. Today, we seek it from algorithms. The questions may be timeless, but the answers—and the responsibility for interpreting them—remain ours.
This article is based on reporting by Phys.org. Read the original article.
Originally published on phys.org
