The findings, and when the research happened
Barna reported on May 19 that 48% of practicing Christians would completely or somewhat trust AI with spiritual growth, compared with 12% of pastors. In a separate question, 34% of practicing Christians agreed that AI spiritual advice was as trustworthy as a pastor’s; 60% disagreed. Concern accompanied that openness: 83% of practicing Christians worried about AI misinterpreting Scripture. These findings come from US research, not a worldwide survey of Christians.
The adult survey involved 1,514 respondents in November 2025; the separate Protestant pastor survey involved 442 respondents in December. Both were online surveys using demographic or institutional quotas. Barna identifies the series as a partnership with Gloo. The public article does not supply the practicing-Christian subgroup size or its operational definition. Read the findings and methodology: Barna: spiritual authority survey and methodology
We are examining this earlier evidence because it supplies a concrete starting point for pastoral discussion. The publication date of this report should not be mistaken for the date people answered the questions.
Reading an attitude without inventing a behavior
There is a consequential difference between being willing to trust advice and having acted on it. A question about openness cannot, by itself, establish how often respondents used a chatbot, which products they consulted, or whether they accepted a particular answer. It also cannot show that a congregation has lost members to an AI service. Those are different research questions requiring different evidence.
The phrase spiritual growth also needs careful handling. One respondent might imagine help locating a passage; another might imagine ongoing personal guidance. Without a fuller account of what respondents understood, a single percentage should not carry an entire theory of religious change. The useful conclusion is narrower: churches have reason to ask what people mean when they describe a tool as helpful or trustworthy.
A comparison between groups does not explain why their answers differ. Training, experience, expectations and the way questions are understood could all be worth investigating; this report does not identify a cause. Nor should a pastor apply a national percentage mechanically to the people sitting in one congregation. Local conversations can illuminate experience, but a handful of conversations should not then be presented as a representative survey.
Trust can mean assistance—or authority
The questions about spiritual growth and pastoral advice should not be collapsed into one result. A person can regard a tool as useful for learning while declining to grant it the authority of a spiritual adviser. Equally, expressing concern about possible errors does not tell us whether that person stops using it. The distinction is worth exploring, without claiming that these published percentages trace individual respondents’ behavior.
A practical church discussion could distinguish three hypothetical uses. Someone asks for the location of a biblical passage; someone requests a comparison of interpretations; someone asks what God wants them to do in a painful personal situation. The first can be checked against a text. The second requires attention to sources and traditions. The third adds personal context and a claim to guidance that a confident answer cannot validate. These are teaching examples, not categories measured by Barna.
This is also why a pastor’s lower reported trust should not become a reason to dismiss a member’s experience. Ask what the tool helped with and where the person encountered uncertainty. A useful response might be to verify a reference together or identify an interpretation the answer omitted. The aim is to strengthen discernment and access to responsible care, rather than win a competition over who sounds more reassuring.
Christian perspective: a helpful answer is not the whole of care
Christian pastoral responsibility includes more than producing plausible religious sentences. In John 10, Jesus describes the good shepherd through knowledge, commitment and self-giving. The passage concerns Christ, not a modern software specification. As a theological starting point, however, it helps explain why Christians should ask what kind of relationship stands behind advice. Read John 10: John 10 (NIV)
A tool may help someone formulate a difficult question or locate material to discuss. That assistance does not settle whether its guidance is faithful, whether the person has been understood, or who will remain accountable if the advice harms a relationship. A church should evaluate those questions explicitly rather than letting the ease of receiving an answer determine its authority.
Acts 17:11 depicts the Bereans examining the Scriptures as they received teaching. This offers a practical principle of testing rather than passive acceptance. Applied to AI, the first step is to open the cited passage and read its surrounding argument. The next is to consider how a proposed interpretation relates to the broader biblical witness and to responsible teaching within the reader’s tradition. A reference that exists may still have been used badly. Read Acts 17:11: Acts 17:11 (NIV)
Human leaders need the same accountability. Comparing a chatbot with an idealized pastor would hide the real possibility of careless or abusive human advice. Christian discernment should ask whether either source can be questioned, whether vulnerable people are heard and whether error is acknowledged. Respect for pastoral ministry is strengthened by responsible practice, not by claiming that religious office makes every answer reliable.
A conversation a congregation can actually have
Consider a hypothetical church member who asks a chatbot how to forgive a relative. The output includes a Bible reference and a confident instruction. A useful pastoral conversation would begin with what the member wanted, what the answer said and which assumptions it made. It should leave room for disagreement and personal context, without demanding disclosure of private information to a group.
A teaching session can use a fictional scenario instead of anyone’s real correspondence. Participants can separate the biblical passage, the interpretation and the proposed action. They can then identify which parts require more information or a conversation with a trusted person. The exercise develops judgment without presenting one impressive or absurd output as a test of every AI system.
Leaders can also explain their own boundaries: which tasks may use AI assistance, what information must stay private and who answers for material shared in the church’s name. Those boundaries should be understandable to ordinary members. A policy hidden in technical language will do little to help someone decide where to take a difficult question.
The pastoral opportunity is to make thoughtful questioning welcome. A person should be able to discuss an AI answer without being mocked for using technology or pressured to accept it. That kind of conversation can reveal what people actually need: information, reassurance, companionship, interpretation or help making a decision. Each need deserves more care than a broad claim that AI is either becoming everyone’s spiritual authority or has no place in religious life.