New NPR/Ipsos poll finds human trust still outweighs AI spiritual advice

The findings behind the headline

AI has entered some Americans’ spiritual lives, but the latest NPR/Ipsos release does not show that machines have displaced human counsel. Ipsos reported on September 28 that 61% of U.S. adults had used an AI tool. Within that AI-using group, 11% had sought spiritual or religious guidance, 13% had generated prayers and 15% had studied religious texts. Those are percentages of AI users, not percentages of all Americans. Source: ipsos.com

Among AI users, 75% trusted advice from people they knew more than AI advice; 6% preferred AI advice and 17% trusted both equally. The poll used the probability-based KnowledgePanel and surveyed 1,011 adults in English on August 7–9. Its reported margin of sampling error for the full sample was 3.6 percentage points; subgroup uncertainty is larger. The release is new, but the interviews happened in August. This distinction matters when connecting the findings to products announced later.

Three activities that should not be treated as one

Looking up a passage, generating a prayer and seeking spiritual direction are different acts. A reader might ask software to locate every occurrence of a word and still reject its advice about what to believe. Another might request a prayer article to adapt personally. A third might use a conversational system to decide whether to leave a religious community. Putting those activities under one heading hides the differences in authority and responsibility.

The questionnaire is therefore more useful than a broad claim that people are turning to AI for religion. Its separate questions allow readers to examine what respondents were actually asked. The published topline identifies the bases for individual results, rather than giving one percentage that can be carried across the entire survey. Source: ipsos.com

For a local congregation, the appropriate response is inquiry. Ask members what they use, for which tasks, and where they go when an answer troubles them. Do not assume a national percentage describes the youth group, the retirement fellowship or the people seeking confidential pastoral support. Each setting can differ from the national adult population.

Availability is not the same as authority

An always-accessible tool may fill a practical gap without earning a person’s deepest confidence. Consider a hypothetical reader studying late at night. They might ask a chatbot for an explanation because no teacher is available, then check a commentary or speak with a pastor the next day. Counting that encounter as a replacement of religious leadership would exaggerate what happened.

The opposite error is to dismiss the encounter because the person still says they trust humans more. A mistaken answer can influence someone before they verify it. A church’s useful response would include clear routes to reliable resources and a realistic way to ask follow-up questions. Telling people simply to stop using a tool may leave the need that prompted them to use it entirely untouched.

Leaders can begin with ordinary provisions: a short list of recommended study resources, a named person to receive difficult questions, and an explanation of how quickly someone can expect a response. These are practical proposals from AI Faith Monitor, not interventions tested by this survey. Their value would need to be assessed in the actual community.

Christian perspective: wisdom has a relational setting

James 3:13–18 connects wisdom with conduct, humility, peace and mercy. It asks readers to consider the character of a life, not merely the elegance of an answer. Applied to AI, that distinction helps explain why religious counsel requires more than plausible sentences. An output can be inspected for accuracy; the system does not thereby become a member of a community who lives out its counsel and accepts correction. James 3:13-18 (NIV)

This does not mean every machine-assisted explanation is spiritually empty. An accurate index or useful translation comparison can help a person attend more carefully to a text. The question is what place the tool occupies. Does it direct readers back to Scripture, responsible teachers and lived obligations, or encourage them to treat a convenient answer as the final word?

A family could use the findings to begin a conversation without embarrassment or interrogation. Ask for an example of an AI answer that helped, then examine its sources together. Ask what would happen if the answer were wrong. The aim is to strengthen judgment and relationships, not to make people hide ordinary questions about faith.

What this survey cannot settle

A poll of reported behavior does not establish whether any particular answer was sound, whether respondents became more faithful, or whether a congregation lost trust in its leaders. It also cannot show a change over time without comparable earlier measurements. This release should not be presented as proof of either an AI religious revival or a collapse of human ministry.

Our earlier Barna report concerns a different study with its own population and questions. The two should be read alongside their methods rather than combined into an invented trend. For researchers, that restraint is essential. For pastors, it preserves a more constructive task: understand the people actually in front of you, learn where they are seeking help, and make trustworthy human guidance easier to reach.