Religious chatbots and scripture interfaces

Religious chatbots and scripture interfaces provide conversational and practice-oriented access to religious material. They can retrieve a passage, compose an explanation, respond to a personal concern or speak through a sacred persona. The distinction matters because locating a text, interpreting it and presenting an answer as religious guidance place different demands on sources and authority.

Retrieval, generation and represented speakers

Catholic source selection: Magisterium AI

Magisterium AI’s May 2026 documentation distinguishes Auto and Magisterial modes, with the latter directing answers toward its selected Catholic source corpus. This is a product’s description of its modes; the service should not be mistaken for an official Vatican voice.^[5] Catholic News Service’s 2024 reporting described linked sources alongside an incorrect pastoral answer and the developer’s acknowledgment that the tool remained imperfect. Providing a source link made an answer easier to inspect without making its application correct.^[6]

Islamic question answering and the fatwa boundary

Muhammad Aurangzeb Ahmad’s taxonomy of Islamic chatbots distinguishes systems by their sources, scope, legal tradition and governance, including the difference between grounded systems and opaque general-purpose services. These distinctions concern which materials and interpretive choices shape an answer.^[2]

Bhatia and colleagues’ Islamic question-answering research tests retrieval and iterative evidence-seeking against a bilingual benchmark that measures free-form errors and abstention. It reports improvements in correctness from grounded and agentic retrieval. That result evaluates the researchers’ answering system under benchmark conditions; it does not authorize it to issue fatwas.^[8]

Egypt’s Dar al-Ifta makes that institutional distinction explicit. Its December 2025 ruling permits AI technology in principle but rejects public reliance on current applications for fatwas, citing the scholarly formation and contextual judgment required of a mufti.^[9]

Recitation checking: Tarteel

Tarteel illustrates a scripture interface whose task is to listen to recitation rather than generate religious advice. Religion News Service’s March 2026 reporting describes medical student and Quran reciter Besher Jabri using it in his practice, alongside an account of peer checking and correction during communal recitation.^[19] The company’s September 2026 support documentation explains that its recognition system can both flag correctly recited words and miss mistakes. It distinguishes optional diacritic detection from tajweed, the rules of Quranic recitation, which it says the system does not check. This is assistance with practice, not certification of a reciter or replacement of a teacher.^[20]

Religious assumptions in general-purpose answers

Two May 2026 preprints test different aspects of religious guidance. Israelsen and colleagues examine asymmetries in hypothetical advice about religious transitions; Wingate and colleagues examine omission of religious considerations from answers to everyday ethical questions. Their author lists overlap substantially, so they belong to a shared research lineage rather than two independent replications. The studies concern benchmarked model responses, not observed conversion decisions or a general measure of religious users’ experience.^[3][4]

Care and institutional deployment

Wester and colleagues worked with 18 chaplains who built or reflected on conversational-care systems. Their accounts emphasize listening, connection, responsibility and restraint. The research examines professional expectations and design judgments; it does not establish outcomes for patients receiving AI care.^[10]

A modified Delphi study by Winiger and colleagues adds a different kind of professional evidence. It surveyed 102 experts in a first round and 83 in a second, finding greater confidence in administrative and informational assistance than direct relational or ritual work. These are expert judgments about possible uses, not patient outcomes; the study did not include patients’ perspectives and was concentrated in North America and Europe.^[16]

The Seventh-day Adventist Church’s South American launch report announced the beta of 7chat.ai from 1 May 2025 and directed users toward official sources and leaders for verification. Salvatierra’s September 2026 article analyzes its publicly documented institutional and theological framing. Because that study uses documents rather than user-interaction or outcome data, its arguments about authority and pastoral expectations are interpretations of the service’s structure, not measurements of effects on believers.^[13][14]

An October 2026 Adventist announcement described the 13 October demonstration of AKAI, drawing on selected biblical and denominational materials, including the Encyclopedia of Seventh-day Adventists and the 28 Fundamental Beliefs. Presenters distinguished access to information from spiritual authority. The report announced an app launch in 2027 while separately anticipating encyclopedia integration the following week. Its descriptions of source-linked answers and safeguards are institutional claims about the system, not an independent audit of its answers.^[17]

Commercial services can offer a broader choice of religious framing. In a June 2025 founder interview, Bible Chat was described as allowing users to select Bible translations and denominations for devotional questions. The interview documents the company’s product account; it does not establish theological accuracy or therapeutic effectiveness.^[18]

BuddhaBot and embodied dialogue

Kyoto University’s BuddhaBot work developed from an earlier non-generative system toward generative dialogue. Its March 2025 announcement describes collaboration with Bhutan’s Central Monastic Body and monastic monitoring before broader access. This documents a particular institutional partnership, rather than authorization by Buddhism as a whole.^[7]

Kyoto’s February 2026 announcement of Buddharoid describes the extension of Buddhist dialogue into a Unitree G1 humanoid. It dates a research presentation to 5 February and discusses ritual assistance as a future possibility. The embodied prototype is treated in Robotic clergy; the announcement does not establish routine ritual deployment.^[15]

Jesus avatars and the setting of conversation

Lucerne’s Deus in Machina placed a Jesus avatar in a chapel’s confessional-style booth in 2024. Organizers explicitly excluded confession and absolution. Its religious setting and represented speaker gave the exchange a different form from a scripture-search tool.^[11]

AP’s 2026 reporting also describes commercial faith-oriented services, including a paid Jesus avatar with memory features. That reporting supplies examples of products and attributed responses, not a census of use. Across these cases, the identity presented by the interface is part of the religious experience being offered; it is not itself proof of institutional endorsement.^[1]

Interfaces representing living teachers, such as Ask Sadhguru and Digital Deepak, present another form of attributed authority. Their relationship to the teacher’s published material is discussed in AI persona spirituality.

What can be checked

The cases locate several different points of accountability: a provider chooses a corpus, researchers define a benchmark, religious personnel oversee a deployment, and a user interprets a represented speaker. A source-linked answer can still misapply a text, while a benchmark improvement says little about a pastoral encounter outside its test. NIST’s generative-AI profile provides general background on confabulation and overreliance; it is not direct evidence about the adoption or religious accuracy of these services.^[6][8][12]

Evidence and uptake

An NPR/Ipsos survey conducted 7–9 August 2026 asked 1,011 US adults about religion and AI. Among respondents who used AI—the question’s unweighted base was 643—15% reported current or past use for scripture study, 13% for generating prayers and 11% for spiritual or religious guidance. These are self-reports about AI generally, not adoption figures for particular religious apps or percentages of all adults. The guidance figure combines 5% current and 6% past use.^[21]

The source set combines product documentation, institutional announcements, direct reporting, practitioner research, benchmark experiments and document-based interpretation. Announced availability is different from sustained use; a governance description is different from an audit of answers; simulated advice is different from a user’s action. The record supports comparison of these functions and cases, but no single adoption rate or uniform judgment of theological accuracy.

See also

• Deus in Machina (AI Jesus)
• AI-generated worship and liturgy
• Institutional religious responses to AI
• Machine-mediated religion
• Hidden context

References

1. Krysta Fauria and Jessie Wardarski, “From ‘BuddhaBot’ to $1.99 chats with AI Jesus, the faith-based tech boom is here.” Associated Press, 10 April 2026.
2. Muhammad Aurangzeb Ahmad, “Islamic Chatbots in the Age of Large Language Models.” Muslims in ML Workshop at NeurIPS 2025.
3. Brett Israelsen, Sheryl Carty, Josh Coates, Nancy Fulda, Julie Park, and Pete Whiting, “When AI Takes Sides on Questions of Faith: Persistent Asymmetries in AI-Mediated Faith Guidance.” arXiv:2605.22975 (2026).
4. David Wingate et al., “Omissive Bias in Religious Representation: Benchmarking LLM Answers to Everyday Ethical Decision-making.” arXiv:2605.24319 (2026).
5. Magisterium AI: “Prompt modes: Auto and Magisterial.” 11 May 2026.
6. Justin McLellan, “Catholic chat bot: Putting AI at the service of the church.” Catholic News Service via USCCB, 26 April 2024.
7. Kyoto University: “BuddhaBot, Buddhist dialogue AI, marks its first overseas release in collaboration with Bhutan's Central Monastic Body.” 3 March 2025.
8. Gagan Bhatia et al., “From RAG to Agentic RAG for Faithful Islamic Question Answering.” Findings of the Association for Computational Linguistics: ACL 2026, 26469–26488.
9. Egypt's Dar al-Ifta: “Using AI applications to obtain fatwas.” 2 December 2025.
10. Joel Wester, Samuel Rhys Cox, Henning Pohl, and Niels van Berkel, “Chaplains' Reflections on the Design and Usage of AI for Conversational Care.” arXiv:2602.04017 (2026).
11. Jamey Keaten, “‘AI Jesus’ avatar tests man's faith in machines and the divine.” Associated Press, 27 November 2024.
12. Chloe Autio et al., Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. NIST AI 600-1 (2024).
13. Cristina Levano, “Igreja Adventista do Sétimo Dia lança novo sistema de inteligência artificial.” Notícias Adventistas, 2 May 2025. Institutional launch report in Portuguese.
14. Marcos Salvatierra, “Religious Authority in the Age of Artificial Intelligence: the Case of the 7chat.ai.” International Journal of Latin American Religions, published 12 September 2026. DOI: 10.1007/s41603-026-00357-5.
15. Kyoto University: announcement of the Buddhist AI humanoid robot Buddharoid. 25 February 2026. Institutional research announcement in Japanese.
16. Fabian Winiger and colleagues, “Artificial Intelligence in Spiritual Care: Modified Delphi Study,” Journal of Medical Internet Research 28, e97094, 27 July 2026. DOI 10.2196/97094.
17. Adventist News Network, “General Conference Announces New Apps Built for OneVoice27,” 15 October 2026.
18. Cate Lawrence, “How Romanian startup Bible Chat turned Christian AI into a global phenomenon,” Tech.eu, 20 June 2025; founder interview.
19. Ulaa Kuziez, “Med student by day, Quran reciter by night: Young ‘huffaz’ carry Ramadan’s most demanding tradition,” Religion News Service, 18 March 2026.
20. Tarteel Help Center, “A mistake was flagged that I didn’t make,” 29 September 2026.
21. NPR/Ipsos, Religion and AI topline, survey 7–9 August 2026, pp. 1, 7–8, Q4(c–e).