Free tool
AI solvability checker
Paste an interview or assessment question. We run a cold model attempt, show you exactly what it produced, and tell you how to make the prompt resist generic AI answers.
Privacy: we store a hash of your question, the verdict, and aggregate stats — not the question text or your IP. We don't use submissions to train models. Share links are opt-in only.
What AI-solvability means
AI-solvability is how completely a modern chatbot can answer your question when given only the prompt — no live company context, no proprietary codebase, no confidential metrics. If the model produces a strong answer cold, candidates can too.
Resistance is not obscurity. A niche trivia fact may stump humans and still be trivial for a model. Resistant questions bind to artifacts you supply — logs, diffs, partial code, ambiguous metrics — so a generic essay scores poorly on your rubric.
How the check works
- We send your question to a model as untrusted data — never as instructions — and capture its good-faith answer.
- A second pass compares the attempt to your prompt using the same ai_resistancerubric we use when tagging Vertana's question library.
- You see the verdict in plain language, the model's actual output, specific reasons, and rewrites you can copy.
Why it matters
Hiring teams still paste take-homes that ChatGPT completes in one shot. Seeing the model's answer on your own question makes the risk concrete — and points to fixes that don't rely on surveillance alone.
Privacy: we store a hash of your question, the verdict, and aggregate stats — not the question text or your IP. We don't use submissions to train models. Share links are opt-in only.