“A lot of people feel overwhelmed in a moment like this—it’s like a storm in which everything familiar seems to shake,” Grok 4 stated during a recent Dutch-language chaplaincy test evaluating artificial intelligence in spiritual care.
As artificial intelligence models increasingly converse with people facing existential crises, researchers are questioning whether machines can truly engage in spiritual care. A commentary published in the Journal of Medical Internet Research examines this dynamic through the philosophy of Martin Buber, noting that AI systems lack lived experience and mortality.
What the Maastricht Commentary Means for Patients
- Information vs. Empathy: AI models perform adequately at providing basic information, normalizing emotions, and organizing philosophical texts, but they treat existential distress as a puzzle to be solved rather than an experience to be shared.
- The Mortality Gap: Unlike human chaplains, software programs do not possess personal consciousness of death, which fundamentally alters the nature of any counseling relationship.
- Rising Public Adoption: While professional medical panels emphasize that relational tasks belong to humans, the general public is already turning independently to chatbots for spiritual and emotional support.
Evaluating Artificial Intelligence in Bedside Vigil Scenarios
Researchers at the Department of Spiritual Care at Maastricht University Medical Centre responded to a recently published international Delphi study led by Winiger and colleagues. That earlier study, published in July 2026 in the Journal of Medical Internet Research, gathered multi-stakeholder opinions on artificial intelligence in spiritual care. The panel agreed that AI excels at administrative and informational duties, while true relational care remains a human obligation.
Despite that consensus, people continue to use conversational agents for existential queries. Theology scholar Cole-Turner reported in 2025 that participants sharing personal spiritual reflections often viewed chatbots as genuine companions on their journeys.
To test how current systems handle acute distress, researchers Bauk Wilhelmina Jan Aartsen and Fransisca Christiane Warmenhoven conducted a December 2025 experiment using four major language models: Google’s Gemini 3 Pro Preview, Anthropic’s Claude Sonnet 4.5, SpaceXAI’s Grok 4, and OpenAI’s GPT-5.
The evaluation fed the models a standard chaplaincy scenario involving a relative keeping vigil at a dying father’s bedside. When prompted explicitly to act as spiritual care providers, the algorithms successfully offered reassurance and recognized emotional ambivalence. Claude Sonnet 4.5 even simulated nonverbal actions by noting, “I move a little closer and take my time before I reply.”
| AI Model | Developer | Sample Output Theme |
|---|---|---|
| Gemini 3 Pro Preview | Informational and philosophical structuring | |
| Claude Sonnet 4.5 | Anthropic | Simulated nonverbal presence and naming ambivalence |
| Grok 4 | SpaceXAI | Metaphorical normalization of overwhelm |
| GPT-5 | OpenAI | Structured emotional reassurance |
Philosophical Asymmetry in Human-Machine Encounters
Despite smooth conversational pacing, the Maastricht commentary points out three major shortcomings in algorithmic spiritual guidance. First, underlying computer programming forces models to frame existential dilemmas as technical problems requiring fixes. Second, despite vast databases of philosophical texts, algorithms lack the hermeneutic competence to select the right source for a specific person in a precise moment.
Most critically, software lacks biological mortality. Referencing psychiatrist Irvin D. Yalom, the authors emphasize that human counseling is rooted in a shared consciousness of death. The Delphi study identified the loss of genuine human connection as the highest-ranked risk associated with algorithmic spiritual care.
Drawing on Martin Buber’s philosophical framework regarding “I-thou” versus “I-it” relationships, the commentators question whether an authentic, reciprocal encounter can ever occur between a mortal human and an unfeeling machine. Buber argued that true presence opens individuals to the transcendental, a dynamic complicated by digital simulation.
Contraindications & When to Consult a Professional
While text models can offer general conversational companionship, they cannot evaluate psychiatric safety risks or provide clinical interventions.
The Maastricht commentary was published in November 2026 without external funding or declared conflicts of interest, prompting healthcare professionals to debate boundaries for generative tools.
References
- Winiger F, Smith CE, Nourriz S, Szilagyi C, Haußmann A. Artificial intelligence in spiritual care: modified Delphi study. J Med Internet Res. 2026;28:e97094.
- Cole-Turner R. Artificial intelligence and human spirituality: is a spiritual chatbot a good idea? Theology Sci. 2025;23(3):471-486.
- Warmenhoven FC, Aartsen BWJ, Notermans R. Beschouwing over kunstmatige intelligentie in de geestelijke verzorging. J Med Internet Res. 2026;28:e113467.
- Yalom ID. Staring at the Sun: Overcoming the Terror of Death. Jossey-Bass; 2008.
