Who programs the digital priest? How AI is changing religious authority

A spiritual question asked at two in the morning no longer has to wait for a place of worship to open. A chatbot on a phone can interpret a sacred text, compose a prayer or produce something resembling pastoral advice within seconds. That convenience raises a new question: when a machine speaks the language of faith, whose authority is actually answering?

Generative AI is entering religious life in two directions. Believers turn to general-purpose chatbots with personal and spiritual questions. Religious groups, meanwhile, build their own databases, specialised models and ethical rules to represent their traditions accurately in the technology’s answers. AI is changing religion, and religions are trying to change AI.

Faith that is always online

A chatbot’s main advantage is availability. It does not get tired, react visibly to an awkward question or require an appointment. A person can ask anonymously about sin, a family crisis, losing faith or the fear of death — subjects that can be difficult to discuss aloud.

But a convincing tone does not establish competence. A model generates text; it has no faith, conscience, spiritual experience or personal accountability. It can mix the teachings of different denominations, invent a source or reduce a difficult question to an excessively confident, definitive answer.

People trust AI — and fear it too

Research suggests an ambivalent relationship. People use AI to find information and understand texts, while being less willing to accept it as a final spiritual authority. Barna Group’s research among Christian users illustrates the difference between accepting a useful tool and handing religious guidance over to it. Pew Research Center’s findings also reveal concerns about the possible weakening of human abilities and social relationships alongside recognition of AI’s capabilities.

That tension makes sense. A system can sound sufficiently human to inspire trust while lacking the qualities on which spiritual authority has historically rested: a lived example, accountability to a community and long-term knowledge of a particular person’s circumstances.

AI already helps prepare sermons

Public experiments have accompanied the religious use of AI. In Germany, hundreds of people attended an experimental Lutheran service whose text and virtual preachers were created with ChatGPT’s involvement. For its organisers, it tested a technological possibility. For critics, it illustrated how easily a formally competent text can appear to substitute for a living relationship.

In everyday practice, AI may help clergy with research, translation, structuring a sermon and administrative work. Risks increase when an assistant becomes an invisible co-author and the congregation no longer knows who selected the arguments, examples and tone.

Religions are trying to change AI too

Religious communities and researchers are not merely users. They develop datasets and evaluation systems to test models on questions involving multiple faiths. AllFaith brings together religious questions across traditions and languages; IslamicLegalBench assesses how well large language models handle complex Islamic legal reasoning. These are research projects, not certifications of a chatbot’s religious authority.

Such projects address an important problem: the texts most commonly represented online are not automatically the most accurate or authoritative. A model that overlooks a smaller language, local tradition or internal doctrinal differences can flatten religious diversity into a single comfortable answer.

There is no neutral digital priest

Training material, system instructions, safety rules, product design and the user’s question all shape an AI answer. A ‘neutral digital priest’ is therefore difficult to conceive. If a model avoids choosing between conflicting doctrines, it may still blur their differences. If it favours one tradition, that choice should be transparent.

The Vatican’s Antiqua et Nova does not equate artificial intelligence with human intelligence. It relates the use of technology to the protection of human dignity, truth and responsibility. The boundary is not merely theological: responsibility for the consequences of a system’s advice remains with people and institutions.

A chatbot conversation is not a protected confession

Spiritual conversations can involve deeply sensitive details about health, family, sexuality, guilt and personal relationships. A conversation with a chatbot should not be assumed to have the same confidentiality protections as a confession or professional consultation. Depending on the service, data may be retained, reviewed for safety or processed in another jurisdiction.

Users should therefore avoid entering identifying details whose disclosure could cause harm. When someone faces self-harm, violence, a severe mental health crisis or an immediate threat, AI cannot replace qualified human help, emergency services or a trusted person.

What does this mean for Georgia?

In Georgia, the question is particularly sensitive. The Orthodox Church plays an important role in public life, while Georgian-language digital religious resources are comparatively limited. A small and uneven body of data can increase the risk that a model imports an interpretation from another tradition or misrepresents the local church’s position when answering a question in Georgian.

Public messages from the Georgian Orthodox Church have also discussed AI as both an opportunity and a spiritual and ethical challenge. A practical response needs to go beyond prohibition: digitising reliable Georgian sources, being transparent about the model and sources in use, setting clear data protection rules and establishing ways to refer people to human support.

AI can serve as a librarian, translator and starting point for a conversation. When it presents itself as a source of final answers, forgiveness or moral decisions, technical speed cannot substitute for authority. The central question is no longer whether a machine can preach. It is who teaches it what truth means — and who is responsible when the digital voice gets it wrong.