The questions you get depend on what you have already said. There is no fixed list and no fixed number — the conversation keeps going until it has a clear enough picture of what you are describing, then stops. That is why your friend finished in six minutes and you took eleven, and why neither of you did anything wrong.
This article explains what "adaptive" actually means, where the idea comes from, and what it does and does not tell you about the person answering. For what the conversation is covering as it goes, see the eight steps of the DISCOVER framework.
The short version
| A fixed questionnaire | An adaptive conversation | |
|---|---|---|
| Question list | Same for everyone | Depends on your previous answers |
| Length | Fixed | Varies — stops when the picture is clear enough |
| If you answer clearly early | You answer the rest anyway | It moves on sooner |
| If your answers are mixed | It has no way to dig further | It asks more in that area |
| Comparable scores? | That's the point of it | Not the point — nothing is scored |
What "adaptive" means in practice
Imagine two students in the first minute.
One says, without hedging, that the thing that holds her attention is taking apart how systems fail — and then says something consistent with that twice more. There is not much value in asking her another six questions circling the same ground. The conversation has what it needs on that dimension and moves to something it knows less about: how she likes to learn, what she wants work to give her.
The other gives answers that pull in different directions — some pointing at working with people, some at working alone with a problem. That is not a worse answer. It is a less resolved one, and it is worth exploring, so the conversation stays in that area a while longer and comes at it from a different angle.
Same tool, two different paths, because the useful next question genuinely differs.
Why it isn't just a longer questionnaire
A fixed questionnaire has to be long enough for its least informative case. Everyone pays that cost, including people whose answers were clear from the start. The trade is between boring people who are already done and under-serving people who need more room.
Adapting removes the trade. The people who need more room get it; the people who don't, don't.
Where the idea comes from
Adjusting what you ask next based on what you have learned is standard practice in assessment design, not something we invented. In computerised adaptive testing, the question of when to stop has its own research literature.
A 2011 paper in Educational and Psychological Measurement by Seung Choi, Matthew Grady and Barbara Dodd compares three approaches to stopping: run until a precision threshold is met, stop when the next item would add little, or cap the number of items. The authors propose a rule based on predicted reduction — estimating in advance how much a further question would actually sharpen the picture, and stopping when the expected gain gets small.
That principle is what "it stops when it has a clear enough picture" means. Not a countdown and not a quota — a judgement about whether asking again would tell it anything.
Why there's no progress bar
A small design decision with a real reason behind it.
You will see a confidence indicator that fills as the picture becomes clearer, rather than a bar telling you that you are 40% through. This is partly because we genuinely do not know how many questions remain — the number depends on your next few answers. But it is also because progress indicators are less straightforwardly helpful than they look.
A meta-analysis of 32 randomised experiments on web surveys by Ana Villar, Mario Callegaro and Yongwei Yang found that a constant progress indicator does not significantly reduce the number of people who give up partway. What mattered was the pace the indicator implied: indicators that moved quickly at first and then slowed reduced drop-off, while ones that started slow increased it.
The lesson we took from that is not to fake a fast bar. It is that a signal about how far along the thinking is is more honest, and less discouraging, than a signal about how much labour remains.
What the variation does and doesn't mean
This is where students most often worry, so let's be direct.
A longer conversation does not mean your answers were worse. It usually means they were mixed, and mixed at sixteen is entirely ordinary — often it means you have a wider range than the person who answered in six minutes.
A shorter one does not mean you were superficial. It means you were consistent.
Nothing is timed and nothing is scored. Taking a minute over a question does not count against you. Neither does changing your mind.
There is no pattern that produces a better outcome. There is no way to answer "well", because nothing is being marked. Answering as accurately as you can produces a description that is actually about you, which is the only thing that makes the output worth anything.
When the picture doesn't come clear
Sometimes it doesn't, and the honest thing is to say so.
Some students genuinely do not have a strong pattern yet. That is normal at sixteen and it is not a failure of the student or of the conversation. When that happens, the right behaviour is to say so and explain what would help narrow things down — not to manufacture a confident top three that isn't really there.
A tool that always produces three neat recommendations is not being more useful than one that occasionally says "there isn't enough here yet." It is being less honest. If you finish and are told the picture is still open, you have been given accurate information about where you are, which is more than a fabricated answer would give you.
Illustrative example — Ananya, Class 11
Ananya answered the first few questions clearly and consistently, and the conversation moved on quickly to areas it knew less about. By the time she finished, she'd answered nine questions in under seven minutes — noticeably fewer than a friend who'd taken almost twenty. She initially worried this meant her answers were "too simple," but the shorter conversation reflected consistency, not a worse or lazier answer. Her friend's longer conversation reflected genuinely mixed signals worth exploring further — neither length said anything about which of them had a better outcome.
What this can't do
It can't tell you what you'll be good at. It works from what you say about yourself, and there are things about you that have not happened yet. Adapting doesn't make it a measurement — better questions are still questions, and nothing here is scored, diagnosed, or validated as a measuring instrument. It can't see what you don't say: if you describe the person you think you should be, you will get a description of that person back. And it can't replace the people who know you — a teacher who has watched you work for two years has information the conversation cannot reach.
What to do next
- Answer as accurately as you can, not as impressively as you can. There is nothing to optimise for and the output is only worth what the input was.
- Don't compare your question count with anyone. It is not a measure of anything.
- Take whatever comes out to a person who knows you and ask whether it sounds like you. Their disagreement is often the most useful part. What's actually in the report sets out what you'll be showing them.
If you want to see the framework this feeds into, our explainer on the eight DISCOVER steps covers what the conversation is actually building toward.
DISCOVER is an AI career discovery experience offered by universities to help students explore what they might study and why. It is designed for guidance and exploration, and does not claim to diagnose, test, or predict.
Related reading
Questions worth asking your child (that aren't "what do you want to be")
"What do you want to be?" produces a shrug. Here are ten questions that ask about evidence your teenager already has, instead.
Strengths aren't the same as marks — here's the difference
A mark measures one task under one set of conditions. A strength is something you do well repeatedly, across situations. Here's the gap between them, and how to find what's on your side of it.
What's actually in a DISCOVER report
Six things go into a DISCOVER report — interests, reasoned career directions, programme fit, a roadmap, and more. Here's what each part is for, and what none of it can tell you.