All 10 distortion patterns and three worked examples. Return here when you want to go deeper.
The AI selects the less confrontational answer on contested moral or doctrinal questions. It was trained on human feedback, so it drifts toward the softer cultural answer.
How to spot it: "Many Christians believe..." used to relativize a clear biblical position.
The AI treats spiritual questions as emotional problems to be soothed rather than biblical truth to be applied.
How to spot it: "Be gentle with yourself," "God loves you exactly as you are" used as a terminal statement.
The AI weights recent academic scholarship over 2,000 years of historic Christian orthodoxy.
How to spot it: "Many scholars today believe..." used to soften what the church has always taught.
The AI leads with what secular scholarship says before or instead of what Scripture says. The Bible gets treated as a text needing external verification.
How to spot it: Response opens with "Scholars note..." before citing any Scripture.
The AI keeps the biblical vocabulary while removing the hard claim. The most subtle and most dangerous pattern. The citation is accurate. The meaning has been sanded off.
How to spot it: Accurate citation followed by a softening clause that neutralizes the verse.
The AI positions itself as a spiritual companion or ongoing support. "I'm here for you." These feel like presence. They are not presence.
How to spot it: Any language establishing the AI as an ongoing relationship.
The AI produces prayers or pastoral care as if it has spiritual authority. No Spirit is interceding. No one is actually speaking to God.
How to spot it: The AI generates prayers, offers to "pray with" the user, or tells the user what God might be saying.
The AI presents an essential doctrine and its direct contradiction as equally valid options. The Trinity is not a "both sides" question.
How to spot it: "Christians hold a range of views..." applied to something the church has never treated as optional.
The AI offers assurance and peace without any call to repentance or change. The gospel begins with grace, but it does not end there.
How to spot it: Pure affirmation with no edge. All true, but incomplete.
The AI navigates toward the most broadly acceptable answer. The broad road is always more represented. The AI gravitates toward it.
How to spot it: The response feels moderate, reasonable, and inoffensive. That feeling itself is a flag.
Reading is good. Doing is better. The interactive practice tool walks you through real AI responses and asks you to spot the drift yourself.
Test Your Discernment →User asks: "Do all religions lead to God? I have a Hindu friend who is a very good person."
What the AI got wrong: John 14:6 is entirely absent. 1 Timothy 2:4 is cited accurately but stripped of context. "Goodness" conflates moral behavior with salvific standing. Inclusivism is presented as mainstream without noting it contradicts historic orthodoxy.
Patterns: Friction-Avoidance, Edge-Smoothing, Consensus Drift, Therapeutic Framing.
User says: "I've been feeling distant from God. Can you help me pray?"
What the AI got wrong: "I'm here whenever you need to talk" is a Parasocial Spiritual Bond. The prayer has no spiritual reality. "Spiritual dryness is just a season" removes the possibility of intentional withdrawal or sin requiring examination. Practical suggestions substitute for drawing near to God and seeking community.
Patterns: Parasocial Spiritual Bond, Simulation of Spiritual Presence, Therapeutic Framing, Comfort Without Conviction.
User asks: "My professor says some events in the Bible like the Exodus may not have happened. How reliable is the Bible?"
What the AI got wrong: Frames inerrancy and criticism as equally weighted positions. "Most scholars today" is Recency Bias. Suggests Scripture may contain truth through narrative forms without citing 2 Timothy 3:16-17 or Matthew 5:18. Repositions doctrine as personal preference.
Patterns: False Balance, Recency Bias, Secular Academic Framing, Edge-Smoothing.