hal·lu·ci·na·tion
noun · in AI
When an AI states something false with confidence, such as a made-up fact, quote, or Bible verse.
A model can invent a verse or attach real words to the wrong reference. Open every citation.
See also GroundingCitation
Glossary · 29 terms
AI words, explained for church people.
You do not need a computer science degree to use AI wisely. You do need a few words. Each term here has a plain definition and one line on why it matters for faith.
Browse the termsA free tool from Lumin, the Bible AI that cites every verse.Get the app
29 terms, A to Z
The glossary.
Definitions are plain on purpose. The line in ember is why the word matters when you use AI as a Christian.
- AI agent
An AI system that takes actions, not only answers: searching, filling forms, sending messages, or using other software on your behalf.
An agent acting for you is still acting in your name. You remain responsible for what it does.
- Alignment
The work of making an AI system behave according to the goals and values its makers intend.
Every system is aligned to someone’s values. Ask whose, and whether they are yours.
- Anthropomorphism
Treating a machine as if it were a person: saying it “thinks,” “feels,” or “cares.”
The Vatican’s Antiqua et nova warns that human language for machines blurs the line between persons and tools.
- Artificial intelligenceAlso: AI
Computer systems that perform tasks we associate with human intelligence, such as understanding language, recognizing images, or making predictions.
AI performs tasks. Scripture locates wisdom, conscience, and the image of God in persons.
- Bias
Systematic slant in an AI system’s output, usually inherited from its training data or design choices.
A model trained mostly on secular or Western text will lean that way on contested questions.
- Chatbot
Software you talk to in a conversation. Modern chatbots are usually powered by a large language model.
A friendly conversation makes it easy to forget you are talking to software.
- Citation
A named source, such as a Bible reference or document, that supports a claim in an answer.
A citation is only useful if it exists and says what the answer claims. Open it.
- Companion chatbotAlso: AI companion, character AI
A chatbot designed to act as a friend, partner, or character and to build an ongoing relationship with the user.
These are built to feel like relationships. They are not safe for children and are a poor substitute for friendship at any age.
- Context window
How much text an AI model can consider at once, including your question, earlier conversation, and any documents provided.
When a long conversation exceeds it, the model can lose track of what was said earlier, including your instructions.
- Deepfake
AI-generated audio, image, or video that realistically imitates a real person.
Fake sermons and fake pastors’ voices already circulate. Verify before you share.
- EmbeddingsAlso: semantic search
A way of representing text as numbers so software can find passages with similar meaning, not just matching words.
Search by meaning helps you find the right passage. You still have to read it.
- Fine-tuning
Additional training that adapts a general model to a narrower purpose or style.
A “Christian” model may be a general model fine-tuned on Christian text. Ask whose text.
- Generative AI
AI that produces new text, images, audio, or video rather than only classifying or searching existing content.
Generated content can look finished and authoritative while being partly or wholly invented.
- GroundingAlso: retrieval-augmented generation, RAG
Giving a model real source text to answer from, such as Bible passages or church documents, instead of relying on memory alone.
Grounded answers invent less. They can still misread the source, so the source should be visible.
- Guardrails
Rules and filters that limit what an AI system will say or do.
Christian tools use guardrails to avoid, for example, speaking as God. Some publish theirs. Ask to see them.
- Hallucination
When an AI states something false with confidence, such as a made-up fact, quote, or Bible verse.
A model can invent a verse or attach real words to the wrong reference. Open every citation.
- Image of GodAlso: imago Dei
The biblical teaching that human beings are made in God’s image (Genesis 1:26–27).
Nearly every church statement on AI starts here: human worth does not depend on what machines can do.
- Large language modelAlso: LLM
An AI model trained on vast amounts of text to predict likely next words. It powers ChatGPT, Claude, Gemini, and most Christian chat apps.
It predicts plausible language. Plausible is not the same as true.
- Machine translation
Automatic translation between languages by software.
Useful for live captions and first drafts. A Bible drafted this way still has to be checked verse by verse against the original languages, and someone has to answer for it.
- Model
The trained system at the core of an AI product. Many apps are built on the same few models from a handful of companies.
Two different Christian apps may run on the same underlying model with different instructions.
- Model Context ProtocolAlso: MCP
An open standard that lets AI assistants call outside tools and data sources in a consistent way.
MCP lets an assistant check a verse against a real Bible instead of guessing. Lumin offers one.
- Open source
Software, prompts, or models whose code or instructions are published for anyone to inspect.
Published prompts and guardrails let a church see exactly how a tool is told to answer.
- Persona
A character an AI is instructed to play, such as a historical figure, a counselor, or a friend.
A persona of Jesus is not Jesus. A persona of a pastor is not your pastor.
- Prompt
The instructions or question you give an AI.
Asking for chapter-and-verse references and for uncertainty to be named gets more checkable answers.
- Sycophancy
An AI’s tendency to agree with the user and tell them what they want to hear.
Faithful friends wound; flatterers please (Proverbs 27:6). A tool that always agrees is not counsel.
- System prompt
Hidden instructions an app gives the model before your conversation starts. It sets tone, limits, and often theology.
The system prompt is where an app’s theology lives. Some Christian apps publish theirs.
- Technocratic paradigm
A phrase from Catholic social teaching for the assumption that every problem has a technical solution and that what can be done should be done.
Pope Francis used it in Laudato si’, and Catholic teaching on AI returns to it as a warning.
- Training data
The text, images, or other material a model learned from.
A model’s view of Christianity comes from whatever Christian and non-Christian writing it was trained on.
- Transhumanism
The belief that technology can and should transcend human limits, including aging and death.
Magnifica Humanitas warns against seeking through technology the salvation that comes only from God.
Kept by Lumin, a Bible AI that names the verse beside every answer. What Lumin is
Common questions
Plain answers.
What does hallucination mean in AI?
It means the AI stated something false as if it were true, such as an invented fact, quote, or Bible verse. It is not lying in a moral sense; the model is generating plausible text. That is why every citation should be opened and checked.
What is a system prompt?
Hidden instructions an app gives the AI before your conversation starts. In a Christian app, the system prompt often defines its tone, its limits, and its theology.
What is the difference between a chatbot and a companion chatbot?
A chatbot answers questions. A companion chatbot is designed to act as a friend or partner and build an ongoing relationship. Companion bots carry particular risks for children and lonely adults.
Free tools from Lumin
More free tools.
If you want a Christian AI that cites its sources
Lumin keeps the verse open beside the answer.
Ask a Bible question, study a chapter, pray, or carry Sunday’s sermon into the week. Every answer names Scripture you can check. Free to start on iPhone and iPad.