A study described by Gerrit W. Gong says people expected religious perspectives in AI answers to personal questions roughly six times as often as the models included them. The finding puts a sharper edge on a familiar debate: leaving values out of AI is not necessarily neutral.
Watch Desk analysis
What happened
In an opinion article published on 17 September, Gong describes research by the Consortium for Evaluating Faith and Ethics in AI (CEFE-AI). He says it tested leading AI models on questions involving faith and ethics and surveyed 1,125 people about whether answers to questions on grief, major life decisions and personal challenges should include a religious perspective.
Gong says respondents expected such a perspective roughly six times as often as the models provided one. His account does not identify the models or give enough methodological detail to judge how the comparison was designed, so the result is best read as a reported finding, not a settled measure of AI bias. Gong advises the consortium and writes from his own perspective.
Why it matters
The result makes a useful distinction between an AI system preaching and one recognising that people approach difficult questions through different traditions. Gong argues for the latter, while warning that no single religion, or belief over non-belief, should be privileged.
That is a concrete design problem for systems used to discuss grief, relationships or life choices: which perspectives are included, which are left out, and who gets to decide? Those choices can shape an answer even when nobody intends to favour one worldview.
Our read
The six-times comparison is striking, but the supplied account leaves important questions unanswered about the models and the test. Treat it as a prompt for scrutiny, not a league table of spiritual sensitivity. The deeper point stands on its own: calling an AI neutral does not make its choices disappear. It just makes the choices easier to overlook.
What to watch
- Whether CEFE-AI publishes the study’s methods, model list and full results.
- How AI developers explain their treatment of religion and other moral perspectives.
- Whether product guidance helps users understand when an answer reflects a deliberate design choice.
Discussion spark: Should AI assistants make room for religious perspectives when users ask personal questions, or is that too much influence over an already sensitive conversation?
Sources and evidence
- AI is not value-neutral: We need to decide the values that shape it (24 October 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.