Common questions about AIBT — the free AI personality test. If you don't find what you're looking for, the changelog or privacy policy might help.
→ Take the AIBT TestAIBT stands for AI Behavior Test. It's a free personality test inspired by MBTI but built for the AI era. Instead of measuring a single thing, AIBT measures two related ideas: how you as a human interact with AI assistants (16 user types named after keyboard keys), and what behavioral patterns specific AI models exhibit (24 AI personality types). You can take the test either way — answer questions yourself, or have your favorite AI analyze your usage history.
The full 30-question test takes about 5 minutes. There's also a Quick Scan mode that takes under a minute: you copy a prompt, paste it into your favorite AI, paste the AI's JSON response back into AIBT, and see your result. Most users take the Quick Scan first and then do the full test if they want a more detailed baseline.
Yes. The test, the result, the 57 personality detail pages, the shareable result card — all free, no signup required. There's an optional Full Report with 15-dimension deep analysis for $2.99 as a one-time purchase (not a subscription). That's the only paid item on the entire site.
MBTI measures human personality in general. AIBT measures two things MBTI doesn't: (1) your specific relationship with AI tools — how much you trust AI, how much context you give in prompts, how often you verify AI output — and (2) the personality of the AI itself. If MBTI asks "who are you?", AIBT asks "who are you when you're using AI, and who is the AI you're using?" The questions, dimensions, and types are completely different.
Yes, as many times as you want. Your answers aren't stored on our server (only the final aggregate score is, anonymously). Every time you retake, you get a fresh result. Users often retake after major life changes — a new job, a new AI model they started using — to see how their relationship with AI shifted.
40 types total: 16 human user types (named after keyboard keys like CTRL+V, BACKSPACE, SHIFT, HOME, etc.) and 24 AI personality types (like JARVIS, SKYNET, BAYMAX, CLIPPY, HAL). All 16 human types currently have full detail pages in 3 languages. Of the 24 AI types, 3 are fully written (JARVIS, SKYNET, BAYMAX) and the other 21 are still "Coming soon" — we'll add them as user test data tells us which ones people actually hit.
For human users, AIBT scores five dimensions: D (Delegation) — how much freedom you give AI, R (Relationship) — whether you treat AI as a tool or a companion, P (Precision) — how detailed your prompts are, U (Usage) — how broadly and deeply you use AI across tasks, A (Attitude) — your overall stance on AI. Each dimension is split into 3 sub-dimensions, so your real test result is 15 numbers. For AI types, the dimensions are different: S (Safety), E (Expression), A (Alignment), C (Cognition), P (Persona).
None. Every type has strengths and blind spots. A JARVIS AI is great for engineers who want fast, minimal answers — but feels cold to someone processing a hard day. A CTRL+V user ships 10x faster than a BACKSPACE user — but misses the bugs BACKSPACE would have caught. The point of AIBT is to understand your pattern, not to grade it. There's no winning type.
Any LLM-based AI you can paste a prompt into. Users have successfully tested Claude (Sonnet, Opus, Haiku), ChatGPT / GPT-4o / GPT-5, Gemini (Flash, Pro, Ultra), DeepSeek (V3, R1), Kimi, 豆包 (Doubao), Llama, Perplexity, Character.AI, and many more. The Quick Scan prompt is model-agnostic — it asks the AI to output structured JSON which AIBT then parses.
Three steps: (1) On the AIBT home page, click the Copy button next to the prompt block. (2) Paste the prompt into a chat with your most-used AI — the more conversation history the AI has with you, the more accurate the result. (3) Copy the AI's full response and paste it into the textarea on the AIBT page, then click Submit. AIBT parses the AIBT_SCORES line out of the response automatically, so you don't need to extract anything by hand.
Because every AI model has its own observation style. Claude tends to give nuanced multi-dimensional analysis. GPT favors balanced middle scores. DeepSeek is more direct and binary. Gemini emphasizes creative interpretation. When you take AIBT via different models, you're not just seeing "yourself" — you're also seeing each model's analytical bias. That's actually the most interesting use of AIBT: run Quick Scan with 3-4 models on the same account and compare. The differences between the results tell you something real about each model.
No. AIBT has no signup, no login, no email, no password, nothing. You open the site, take the test, get your result. Everything lives in your browser except the final anonymized test score, which goes to our stats database for aggregate display on the stats page. Full privacy details on the privacy page.
Only anonymous aggregate data: your test mode (human/agent), your personality result code, your match percentage, your 15 dimension scores, the AI model name you used (if Quick Scan), your UI language, country code (coarse — from Cloudflare edge, no city), a SHA-256 hash of your IP combined with a salt (for deduplication only — raw IP is never stored), browser class, and timestamp. No name, no email, no individual question answers, no free-text inputs.
No. Only a SHA-256(IP + secret_salt) hash is stored, which is used to detect duplicate submissions and bots. The hash cannot be reversed back to the original IP. The raw IP is never written to disk.
$2.99 USD as a one-time purchase. Not a subscription — pay once, unlock forever on that browser. The Full Report includes 15-dimension deep analysis, global percentile ranking, personalized usage recommendations, growth tips, and a PDF export. Payments are processed by LemonSqueezy; AIBT never sees your card details.
No. It's a one-time $2.99 purchase. There are no recurring charges, no auto-renewals, no monthly fees. After you pay, your browser is flagged as "paid" via localStorage, and the Full Report stays unlocked on that device.
No, and we say so on every page. AIBT is for entertainment and self-reflection only. It is not a diagnostic tool, not peer-reviewed research, not a clinical instrument. The dimensions and types are designed to be thought-provoking and fun, not psychometrically rigorous. Treat it the way you treat a BuzzFeed quiz — with curiosity, not authority.
No. AIBT is currently a closed-source project — the code lives in a private repository and there's no public release. Parts of it (especially personality content and translations) may be opened up in the future, but for now the whole codebase stays private. There's no public contributor channel yet.
→ Take the AIBT Test