SailToNUST checks answer keys before a question goes live. A printed key gets read twice and compared, and a question without a trusted key gets solved again by a solver that never sees the original answer. In November 2026 we ran 4,648 full-paper questions through that blind check, corrected 47 keys and removed 89 questions.
Wrong keys are the most common problem with free NET material. A PDF passes through ten WhatsApp groups, someone retypes it, and option C quietly becomes option D. You practise, you get it "wrong", and you learn the wrong method. This post explains what we do about that and what you can do when something slips through.
SailToNUST publishes this post about its own process.
Check 1: reading a printed key twice
Most preparation material prints an answer key, often as a grid at the end of a chapter. Misreading one cell in that grid would corrupt a whole run of questions, so we never trust one reading.
Two separate tools read every key page, and a script compares the two readings cell by cell. When they agree, the key goes in. When they disagree, that page gets a third, closer reading before anything is accepted. The two tools come from different makers on purpose, because two copies of the same tool tend to make the same mistake and agree on it.
Check 2: solving the question blind
Some material has no printed key, and recalled questions often carry a key someone guessed. Those questions get solved again from scratch.
The solver sees only the question and its options. It never sees the key, because a checker that can see the answer tends to agree with it. Then we compare:
| Result | What happens |
|---|---|
| The blind solve matches the key | The question goes live |
| It disagrees, and a second blind solve sides with the key | The question goes live with its key |
| It disagrees, and the second solve sides with the first | The key is corrected |
| The solves disagree with the key and each other | The question is removed |
A question is also removed if it has two defensible answers, no correct answer or a missing figure. Vocabulary items fail most often, usually because two options mean the same thing.
The numbers
| Batch | Checked | Kept as is | Key corrected | Removed |
|---|---|---|---|---|
| Questions already in full papers | 2,589 | 2,524 | 14 | 51 |
| New questions for full papers | 2,059 | 1,988 | 33 | 38 |
| Total, November 2026 | 4,648 | 4,512 | 47 | 89 |
Removed questions were replaced with checked questions from the same subject, so every paper still has 200 questions at the right split. When we rebuilt the English sections of six older Engineering papers this month, 480 candidate questions went through the same check, and 50 were turned away before any reached a paper.
Check 3: you
No check catches everything. A typo in an option, a unit that went missing or an explanation that skips a step can survive both passes. That is what the Report button on every question is for.
Press it, pick what is wrong and add a note if you like. The report comes straight to us with the question attached. We fix the question in the chapter bank and in every full paper that uses it.
What this means when you practise
Trust the key, but check the method. If your answer differs from ours, work the question again before you move on. Most of the time you will find your slip. If you still think the key is wrong, report it. Either way you learn more than you would by memorising a letter.
You can see the keys and working for yourself in the practice bank, in the review after any full paper, and in the past MCQs. The post on whether SailToNUST MCQs are AI-generated covers where the questions come from.
Key facts
- Printed keys are read by two separate tools and compared before they go live.
- Questions without a trusted key are solved blind, by a solver that never sees the key.
- In November 2026, 4,648 full-paper questions were checked: 47 keys corrected, 89 questions removed.
- The Report button on every question sends it straight to us.
Source: SailToNUST answer-key audit records, November 2026




