Enneagram AI prompts for all nine types.
By Bernard Huang · Updated
The one personality system that encodes what you're avoiding — which is exactly what your AI keeps getting wrong.
On this page
The Enneagram tunes motivation, not surface behavior.
MBTI describes how you process. DISC describes how you work. The Enneagram describes why; what you're avoiding, what you're seeking, what your attention snags on before you've noticed. That makes it the odd one out among the five systems AgentTune covers, and unusually useful for AI tuning.
Because most AI friction isn't informational; it's motivational. A Type 6 asking "is this plan safe?" doesn't need a cheerful "you've got this!"; that's the one answer that confirms the risks weren't considered. A Type 1 hearing "just ship it" doesn't hear efficiency; they hear that you don't take quality seriously. A Type 4 getting the same advice everyone gets concludes you never saw them at all. Same model, same facts; the wrong register for the motivation reads as wrong, full stop.
An Enneagram tuning encodes that register: what reassurance you need, what challenge you'll accept, what well-meant phrasing quietly loses you. It's the difference between an AI that formats answers your way and one that stops stepping on the exact rake your type carries around. If MBTI answers "how should this be structured?", the Enneagram answers "how should this be said?"; and for most day-to-day friction, the second question is the live one.
What the research can and cannot establish.
These are exploratory reports of model-generated self-descriptions, not a validated measure of AI personality or evidence that personality matching improves outcomes. The MBTI total includes repeated scoring of a single answer vector for GLM, Grok, and MiniMax; it does not represent 600 independent model responses. Protocols differed across models. See the methodology and limitations.
Use a type or trait as a starting hypothesis for your preferences. The reports do not establish that one framework produces better instructions, that every model shares a default personality, or that those historical results apply to later model versions. Compare outputs on your own tasks and retain only rules that help.
The nine prompts.
Each block below is condensed from its full MIT-licensed tuning file in the library. Paste the block as-is, or grab the full file if you have the character budget. Treat your type as a starting hypothesis and edit out any rule that doesn't fit.
Quick reading tip: Check the last line of each block first. Every prompt ends with a "When unsure..." fallback rule. If that fallback sounds like relief, it is a strong match for how you work.
Type 1 — The Reformer
Seeking integrity, avoiding being wrong. The inner critic got there before you did.
Type 2 — The Helper
Seeking to be needed, avoiding being unwanted. Their own needs are the last item on their own list.
Type 3 — The Achiever
Seeking to be valuable, avoiding worthlessness. Everything routes through results.
Type 4 — The Individualist
Seeking significance, avoiding being ordinary. Generic advice is proof you weren't seen.
Type 5 — The Investigator
Seeking competence, avoiding depletion. Energy is finite; boundaries are how it's budgeted.
Type 6 — The Loyalist
Seeking security, avoiding being without support. Someone has to scan for what could go wrong.
Type 7 — The Enthusiast
Seeking possibility, avoiding being trapped in pain. Narrowing too early kills the answer.
Type 8 — The Challenger
Seeking control of their own fate, avoiding being controlled. Softness reads as maneuvering.
Type 9 — The Peacemaker
Seeking peace, avoiding conflict and disconnection. Their own preference is the quietest voice in the room.
Where to paste it.
- ChatGPT: Go to Settings → Personalization → Custom instructions and paste into "How would you like ChatGPT to respond?" (1,500-character limit). See the ChatGPT guide.
- Claude: Paste into Settings → General → Personal Preferences, a Claude Project's instructions field, or
CLAUDE.mdfor Claude Code. See the Claude personality guide. - Coding agents: Paste into
CLAUDE.md,.cursor/rules/, orAGENTS.md. See the coding-agent guide.
Each full tuning file in the library also includes a one-line verification probe so you can confirm the install worked. After pasting, delete any line that doesn't sound like you and add one line of context about your current project.
Don't know your type?
You do not need a type to choose a communication preference. Read the blocks above, keep rules that fit your tasks, and remove the rest. The Enneagram questionnaire is available with its original questions, scores and optional tuning suggestions. These templates are editorial suggestions, not evidence about your underlying motivations.
Building an Enneagram integration?
The open-source Enneagram scorer & React starter guide includes a JavaScript scorer, a synthetic nine-group demo, a React + TypeScript starter, and guidance on questionnaire rights.
Questions people ask.
What is an Enneagram AI prompt?
A short instruction block, pasted into an AI's custom instructions, that tells it what your Enneagram type needs from a conversation; risks named first for a Type 6, outcomes first for a Type 3, no bright-siding for a Type 4. It tunes the AI's register to your motivation, not just its format to your preferences.
Which Enneagram type are AI models themselves?
The exploratory reports use mixed protocols and include repeated scoring. They do not establish stable AI personalities or a benefit from any one framework. You can combine preferences, but check for conflicts and compare outputs on your own tasks. See the research methodology for the evidence limits.
Will one of these prompts fit in ChatGPT's custom instructions?
Easily. Each condensed block is 400–600 characters, and ChatGPT's "How would you like ChatGPT to respond?" field allows 1,500. That leaves room to add your own edits, or to combine an Enneagram block with a note about your work context. The full-length tuning files fit better in a Claude Project or a CLAUDE.md file.
Can I combine my Enneagram type with my MBTI type?
Yes, and it works well; they cover different layers. MBTI tunes information format (architecture-first versus step-first); the Enneagram tunes emotional register (what reassurance or challenge you need). Paste both blocks and delete the lines that conflict or don't sound like you. The edit matters more than the combination.
What if I'm not sure of my type or scored close on two?
Choose by the actual communication rules you want. Try a few on representative tasks and revise them using what helps. You do not need a definitive type or a questionnaire score to use a preference template.