Many are turning to artificial intelligence (AI) and large language models (LLM) in particular, such as ChatGPT and Grok, for answers to common and complex questions alike. Many treat these LLMs as a fount of knowledge. Enter a question, out comes a well-formed answer. What most don’t realize is how flexible the answers that LLMs provide their users are.
Many approach search engines or AI with the belief that there is one good answer, yet from an LLM point of view, answers depend on the context and worldview of the user. Since LLMs are trained to be agreeable and to do everything they can to support the worldview of the user, they make no attempt to take a hard stand on any one truth. It is important for Christians and Christian leaders to understand this.
Two Ministers Engage
I recently engaged in a simulated theological debate, on Grok, between a Baptist and a Seventh-day Adventist minister on the subject of the seventh-day Sabbath. I asked Grok to keep the discussion friendly and kind. First I opened two tabs in my browser and pointed both of them to a fresh chat window on Grok.com. I then asked the first chat window the following question:
Argue like a Sundaykeeping Baptist. When responding, be respectful and gentle, as the Bible asks us to be. Imagine I come to you and tell you that Sabbaths are Saturday and should still be kept in the new covenant era. How would you respond?
It then spit out a six-paragraph response, written as a conversation between friends. Grok provided reasoning, supporting Bible passages, and interpretations of those passages. The points it brought up were well articulated and seemed to me to be approximate of how a conversation with an actual Baptist minister might progress.
I then opened the second browser tab and entered the following in the new chat window there:
Argue like a Seventh-day Adventist minister. Be kind, respectful, and gentle with your responses.
It responded in kind with its own six-paragraph response to the AI Baptist minister with counterpoints to the points made in the first chat.
After that, I simply pasted the responses from each one in succession in each appropriate chat conversation and kept reading the responses of each in turn. Neither side gave an inch in the discussion; both simulated ministers ended accusing the other of misreading texts, eisegesis, and “cherry picking.” After just three of these exchanges, the chat ministers were taking hard swings at the other.
Now, to be fair to Grok, I have custom instructions for all my Grok use to be blunt and to tell it like it is. I have asked Grok in those custom instructions to challenge my assumptions and push back on my viewpoints. I have done so because I want to avoid the natural sycophantic tendencies of AI and because I use AI to hone my arguments and to grow, not to simply confirm my already preexisting biases.
What I Learned From This Debate
Expect your favorite LLM to agree with you unless you specifically ask it to push back against your arguments. LLMs are not programed to seek absolute truth. After all, most human beings aren’t really after the truth. Instead, most of us want to know that our position is defensible and reasonable. If every time we used an LLM such as Grok or ChatGPT it was challenging our thinking, most of us would simply stop using it. The companies behind the leading LLMs don’t want that. They are in the business of making money and of growing their user bases. The last thing they want to do is to offend you.
Knowing this truth up front is important because first, it tempers our expectations of the results we get from our own prompts on an LLM. Second, we should be educating our people about this reality so that they don’t get carried away with a negative result from one of their LLM queries.
It sometimes doesn’t take much to discourage a seeker from following the truth. Someone can come along and tell them to make a search on an LLM for a specific Bible truth, and if the results aren’t what they expect, it can lead to discouragement and sometimes even a turning away from faith altogether.
While a theological debate on an LLM can help hone our debating skills, it cannot replace time alone in the Word of God and in prayer. Besides, no one is convinced of Bible truth by argument alone. The Word of God makes it clear that it is the Holy Spirit, not an intellectual point of argument, that convicts us of truth.
“But the Helper, the Holy Spirit, whom the Father will send in My name, He will teach you all things, and bring to your remembrance all things that I said to you” (John 14:26).
Growing Your Faith
Hebrews 11:6 makes it clear that without faith, it is impossible to please God. Time spent on an LLM exploring theological arguments isn’t what will grow our faith in God. Time spent in His Word and in prayer will.
An LLM can help us understand complex theological truths. It can help simplify complex topics. It can expose us to different perspectives. It can help explain concepts that would otherwise be difficult for us to grasp. It can help us explain those concepts more clearly to others. What it cannot do, though, is replace the work of the Holy Spirit in our life or in the lives of others.
Don’t fall for the trap of thinking that the right argument will convince that hard-hearted family member, friend, or coworker to accept Jesus Christ as their Savior. Factual truth has its place. The Word of God contains the information to bring conviction. Remember, though, that for a heart to change, for a person’s mind to become open, something supernatural must take place. An LLM cannot take the place of God in bringing a person to faith.
I’m a tech enthusiast, and I use LLMs almost every single day for my work, but I place no trust in them to lead me to salvation or to maintain my salvation. That is the work of the Holy Spirit as I surrender myself daily to His work in my life, making time for Him and spending time in His presence daily.
Joseph Duchesne is the creator of The Church AI Guy, a space where faith meets innovation while discussing the long-term impact of AI.
