See how it works
A guided tour through every step — from asking your question to recording what you experienced.
Ask your question naturally
Type the question exactly as you would normally ask it — to a friend, a search engine, or a doctor. Don't try to make it more precise. The whole point is to capture real human questions, not polished ones.
Your question can be long or short, specific or vague, personal or abstract. All of these are valid research contributions.
Choose the closest interpretation
Before generating an answer, the AI proposes several ways a person might read your question and several reasons someone might ask it.
You review those possibilities and choose the ones closest to what you meant. You remain the authority on your question.
You can also rephrase the question if a different wording would better capture what you meant.
Read the answer and what was considered
The AI answers using the interpretation and reason you selected. You also see three cards showing:
Your selected meaning — the interpretation you chose, plus the alternatives you were offered.
Your stated reason — why you said you asked, and the other options.
The explanatory frame — the approach used for the answer, why it fits your selection, and what could go wrong if the selection did not match your meaning.
Decide what to do next
After reading the answer, you have three choices. This is where you take control of the research process.
A — Accept the answer and move on to recording your evaluation. You were satisfied with what you got.
B — Revisit your interpretation. Maybe you picked the wrong meaning, or you want to see how the answer changes with a different selection. Your radio buttons are still set — just adjust them and get a new answer.
C — Rephrase the question. Start fresh with a reworded version of your question. Your original wording is pre-filled in the question box, ready to edit.
Tell us what you actually meant
This is the most important part. You record your own version of events: what you actually meant, why you really asked, and how the answer landed.
You'll also use 0–100 sliders to rate how closely the selected meaning and reason matched, how useful the answer was, and how difficult it was to understand.
You can update this evaluation later — every version is preserved, so your change of mind is itself part of the data. And you can revisit past questions to try perturbations: the same question with a different interpretation, or slightly different wording.