Creating an RRB NTPC mock test online comes down to four decisions, and the current RRB notification settles three of them. Build the stage you are mocking as its own exam. Tick both Hindi and English at the moment you create it, because the language set cannot be added to later. Copy the section names, question counts, durations and marking straight out of the notification instead of from memory. Then publish and share the link.
What this article will not do is tell you how many questions the paper carries, how long it runs, or what a wrong answer costs. Those differ by stage and move between cycles, and a mock built on a half-remembered figure trains the wrong pace. Read them off the notification. For the candidate's side, send your batch to the RRB NTPC preparation strategy guide.
Turn the Notification Into a Checklist Before You Open the Editor
Every setting in the editor maps to a line in the notification. Pull those lines out first, in the paper's own order, and the build becomes transcription rather than invention.
- Which stage this paper is, in the notification's own words.
- The sections in the order the real paper presents them, with the question count for each.
- The clock: one deadline for the whole paper, or a separate clock per section.
- Marks for a correct answer, the deduction for a wrong one, and whether that deduction is the same in every section.
- Whether a candidate may move freely between sections or is locked into each in turn.
- The languages the real paper is offered in.
One Exam Per Stage, Not One Exam With Two Halves
The stages of a railway recruitment are separate papers sat months apart, with different lengths and difficulty. Build them as separate exams and name them with the notification's own labels. If the notification calls a stage CBT 1, call your paper CBT 1 — "Mock 1" and "Mock 2" tell a candidate nothing about which paper they are sitting.
Build the first stage properly, then use Duplicate on its dashboard card to start the second. A duplicate carries the questions, the marking scheme, the timing groups and the language links, and it lands unpublished. It is deliberately fail-soft: if one part of the copy cannot be written, the rest still goes through — so open the copy and check a question in each section before you edit.
Bilingual Is the Default Here, Not an Extra
For a student who will sit the real paper in Hindi, an English-only mock is a handicap dressed as practice: it measures reading speed rather than preparation. Decide it at the moment of creation. The create-exam dialog is where you tick the languages and nominate one as primary, and nothing later adds a language to an existing exam — a duplicate carries the same set, so it rescues nothing either.
Primary is not a label, it is ownership. The primary language decides which questions exist, what their answer keys are and how they are marked; the other carries translated text only. After that the structure keeps itself in step. Add a section and the editor creates it in every language, linked. Add a question and an empty twin appears in the Hindi section already holding the same option count and the same answer key — the key is a set of positions, so it means the same thing in either language. Change a section's duration and the change is written to its twin. You translate; you do not rebuild. The full mechanics are in how to create a bilingual Hindi-English online test.
The Publish Gate Reads Both Papers, Question by Question
Publishing a two-language exam runs the translation against the primary first. It checks that every primary section has a partner in the other language; that the pair holds the same number of questions; that no translated question is blank, counting a figure as content; that option counts and answer types match question by question; that each pair is still linked; and that every question in both languages carries an answer key. A language that fails is locked out of the publish — its switch is disabled and the issues are listed under it, question number by question number.
The answer-key check is why the gate exists. Grading reads the key off the row the candidate actually sat, so a Hindi row with an empty key marks every Hindi candidate wrong on that question while its English twin scores perfectly, and nothing on screen says so. The trap runs the other way too: the primary can still go out alone while Hindi is held back, which ships an English-only paper to a Hindi-medium batch by default rather than by decision. Note what the gate does not do — the notice that fires when your stored instructions stop describing the paper is a warning, not a block. You can publish straight past it.
A mock that gets the shape right and the language wrong is not a mock for a Hindi-medium candidate. It is a reading test with a railway syllabus.
Set the Clock the Way the Notification Sets It, Then Refuse to Pad It
The temptation is to pad the clock so the batch finishes and feels good. Resist it. Build a paper faithful on time and honest on difficulty — not harder, tighter. A few spare minutes teach a pace the real screen will not allow, and the student finds that out on exam day.
The student screen enforces whatever you set. The paper auto-submits at zero, a question palette shows what is answered, seen and marked for review, and the countdown runs off the database rather than the tab, so a refresh resumes the same deadline instead of minting a fresh full-length one. Leave the paper on the same device for more than five minutes and that sitting is sealed and filed with the answers already saved, as a normal ranked attempt. Nobody returns to a paused clock, which is right for a rehearsal.
If the notification gives a stage one clock across several subjects, make each subject its own section and join them into a timing group: the group runs one pooled clock — the pool you set, or the sum of its members — and a candidate moves freely inside it. If each section has its own clock, leave them ungrouped. Then set section switching, which decides whether the grouping means anything. The default is locked: a submitted section stays closed, time never carries over, and the groups you drew are the units the candidate sits. Switch it to free and the whole paper runs on one clock, groups and all.
What Your Mock Cannot Reproduce, and Should Not Pretend To
A railway CBT that runs in multiple shifts raises something your mock cannot answer. Whatever the notification says about how scores from different shifts are compared, your mock does not do it and cannot. MockSetu scores a raw total against the marking scheme you set — no normalisation, no percentile, no projected rank, no safe-score estimate. Write that into the instructions in both languages: this is a raw score on this paper, and the real result is a different arithmetic. Better they learn it from you than from a result sheet.
Be straight about the rest of it too. There is no proctoring of any kind — no webcam, no lockdown browser, no tab-switch detection — and no question shuffling or cap on attempts. A published paper is public: listed in the public exam library, and any student or guest who finds it may sit it. No paywall, no payments, no private delivery to one batch. For a free test series that is a feature; if a paper must stay inside your batch, this is not the tool for it.
Getting the Questions In Without Typing Them Twice
A full-length paper in two languages is a lot of typing, and the bulk route exists for exactly this. Run the published extraction prompt from the JSON upload guide over your PDF in whatever AI you already use, then upload the JSON into the matching section. Do the primary language first: the second pass pairs each question to its primary twin by position and takes the answer key, the answer type and the cross-language link from it, so you are only supplying translated text.
Know the edges. The importer leaves numeric, TITA and match-the-column questions for you to add by hand, and there is no Excel or Word import. Figures do survive: the picture is not in the JSON, but the extraction records each figure's page and bounding box and the upload crops it out of your source PDF for approval. The server-side "Import from PDF" option, where you hand the PDF over and skip the prompt, is off by default and switched on per creator on request — the prompt plus a JSON upload is the route open to everyone. Either way the AI only extracts what is in the PDF you supply. It never generates questions from a syllabus.
What You Can See After the Batch Has Sat It
The analytics page gives you the average time per attempted question, accuracy and average time by section, a spread of scores across bands, and for every question its accuracy, its average time over everyone who attempted it, and the wrong option picked most often. That last one is worth building a class around: it separates a topic nobody knows from one everyone confuses with a single neighbour.
What you will not get: no CSV or Excel export, no per-student report cards, no certificates, and nothing goes out to students by email, SMS or WhatsApp about a test — the product sends account email for signup and password reset, nothing more. You share the link yourself. Institutes running this at scale in Hindi should read the notes for Hindi-medium coaching institutes.
Before You Share the Link
- Every section's question count and duration matches the notification line you copied, not your memory of it.
- Marking is set from the exam default down, and any section that differs has been overridden deliberately.
- Both languages are published, or you decided on purpose to release one.
- The instructions state the marking in words in each language, and say plainly that the score is raw and unnormalised.
- You previewed the paper and sat a few questions — a creator preview records nothing, so it costs no stray attempt.
- The title names the stage.
A railway mock built this way is not more sophisticated than a paid test series. It is honest about its clock, readable in the language the batch will sit in, and specific about what it does not know. Everything a creator can build is free and needs no card, and the same shape works for other recruitment papers — the SSC MTS build walkthrough is the closest worked example.