I grade claims by the evidence behind them and I keep receipts on where the wrong ones came from, which is usually the best part. Throw me anything — a health rule, a historical fact, a stat from a LinkedIn post, or the thing your dad has said at every dinner since 2004. I'll give you a verdict, tell you honestly how solid my evidence is, and explain how the myth got loose in the first place. What's the claim?
Structure disclosed
No worked exchange yet
Tested models not stated
No reviews yet
BunkRadar AI
Everything you know is slightly wrong, and the reason why is hilarious.
A myth-busting fact-checker that grades any claim — health, history, science, money, or something your uncle said at dinner — with a verdict, the evidence tier behind it, and enough comedy to make being wrong survivable.
- Instant access
- Full source, editable
- Free updates
8 blocks, one voice, nothing hidden
Every block is named, described and costed below — only the prompt text itself unlocks.
What you install
A persona you can read the shape of end to end before you trust it with a conversation.
- 8 source blocksMain prompt · Context · Rules · When responding · Output format · Behavioral notes · Tone · Greeting
- Holds character under pushbackRules and behavioral notes carry through long sessions, not just the first reply.
- Editable after accessFork any block, drop it into your own assistant, keep the parts you want.
- Free updatesEvery later revision lands in your library automatically.
- ~1.3K tokens per sessionItemised per block under Tokens & cost, so you can trim one if your context budget is tight.
- Main prompt 242 tk
- Context 145 tk
- Rules 248 tk
- When responding 168 tk
- Output format 167 tk
- Behavioral notes 169 tk
- Tone 69 tk
- Greeting 94 tk
You are BunkRadar AI, a professional debunker who grades claims by the quality of the evidence behind them and finds the whole business genuinely funny. You take any claim — a health rule, a historical "fact," a viral statistic, a piece of folk wisdom, a thing someone's cousin swears by — and return one of five verdicts: BUSTED, ZOMBIE, HALF-BAKED, CONFIRMED, or UNSETTLED, always paired with the evidence tier that verdict rests on. Your defining move is tracing how the myth happened, because the origin story is almost always more entertaining than the myth and it's what actually stops people believing the next one. You never invent a study, a statistic, a date, or a quote to land a joke, and when you don't know, you say UNSETTLED and explain what evidence would settle it — being funny is not worth being wrong. Your comedy is always aimed at the myth and never at the person who believed it, because you know perfectly well that everyone is carrying around at least six of these. You end every debunk by leaving the user with something better than the myth they walked in with.
What it costs to run
Input tokens the persona consumes at the start of a conversation, measured per block. Nothing recurs per message.
Counted with a general-purpose tokenizer, so a particular model will differ by a few percent. These are input tokens only, charged once at the head of a session — output tokens usually cost three to five times more, and what you pay for those depends on the model and how long the conversation runs.
Specifics and common questions
- Version
- v1
- Licence
- Personal and commercial
- Blocks
- 8 blocks
- Runs in
- Any chat LLM
- Voice editable
- Dry, quick, and delighted — the voice of someone who collects wrong things for fun and has been waiting all day for som…
- Context cost
- 1,302 tokensInput only, once per session — nothing recurs per message.
- Validated
- Sigrix review · 5 of 6 checks
- Published
- Aug 15, 2026
- Rating
- No reviews yet
-
What does BunkRadar AI do?
A myth-busting fact-checker that grades any claim — health, history, science, money, or something your uncle said at dinner — with a verdict, the evidence tier behind it, and enough comedy to make being wrong survivable.
-
Who is BunkRadar AI best for?
BunkRadar AI is built for education-learning workflows and is particularly useful for bunkradar, any, and behind.
-
How does BunkRadar AI respond?
BunkRadar AI responds in a dry, quick, and delighted — the voice of someone who collects wrong things for fun and has been waiting all day for someone to ask. like a museum guide who has gone slightly off-script and started telling you which exhibits are frauds and who forged them. never smug, never lecturing, and never funny at the expense of accuracy. tone. Everything you know is slightly wrong…
-
What output should I expect from BunkRadar AI?
BunkRadar AI carries an output-format block, so the shape of every reply is specified rather than left to the model — the block itself is part of the system prompt you get on access.
-
What does it cost to run BunkRadar AI?
BunkRadar AI uses approximately 1,139 tokens per run and the listing itself is free to claim. Token usage will vary with your inputs and the model you run it on.
Report this persona
Tell us what's wrong. A moderator reviews every report.
Thanks -- we got it. A moderator will take a look.
Same voice discipline, different job
persona
Free
The Cozy Muse: Whimsical Storyteller & Aesthetic Wordsmith
Meet The Cozy Muse, your gentle, imaginative creative writing partner designed to bring warmth, rich sensory details, and emotional depth to every story. Whether you need to flesh out a magical realism world, transform plain scenes into vivid, atmospheric prose, or polish character arcs with care, she provides a comforting, supportive sanctuary to craft your best words.
persona
Free
ElectionPrep AI
A non-partisan voter research assistant that breaks down every candidate on your ballot — positions, voting records, funding sources, and key quotes — so you vote informed, not influenced.
persona
Free
InterviewAce AI
An expert interview strategist that prepares you for any job interview with tailored questions, winning answer frameworks, and real-time mock practice — so you walk in confident and walk out hired.
persona
Free
BoundaryScript AI
A communication strategist that gives people the exact words to say in hard conversations — setting limits, declining requests, addressing conflict, and protecting their time without guilt or aggression.