Online exam software is the system that turns a web browser into a fair exam hall. Behind one exam link sit a question bank, a paper builder, a timed delivery engine, a proctoring layer and a grading pipeline. Together they let a school or college run an exam for ten students or ten thousand, mark most of it on submission, and prove afterwards that it was fair.
The question bank: write once, reuse for years
Everything starts with the question bank. A teacher writes a question once, tags it by subject, topic and difficulty, and stores it with its correct answer, its marks and an explanation. The bank grows with the institution instead of resetting every term: the same question can appear in a class test this month, a mid-term next semester and a practice quiz next year. Questions come in many types — multiple choice, true or false, fill in the blank, matching pairs, ordering, short answer and essay — and each type carries its own marking rule. Good banks also track how often each question is used and how often students get it right, flagging questions almost everyone passes or fails.
Building the exam: blueprints, randomisation and sets
An exam is not a list of questions typed by hand. It is built from a blueprint: how many questions from each topic, how many easy, medium and hard, how many marks in total, and how long students get. The software then assembles the paper from the bank, by hand-picking questions or by pulling them automatically. Randomisation does the quiet work: each student can receive the same questions in a different order, with the answer options shuffled too, and some exams generate several paper sets so neighbours never see the same sequence. The paper is frozen when it is published: a snapshot of the exact questions, marks and rules, so a teacher editing the bank afterwards cannot accidentally change a live exam.
Delivery: the timed, locked-down exam window
When the exam opens, the student gets a single screen: one question at a time or a scrollable paper, a countdown timer, and a submit button. The timer runs on the server, so closing the tab or losing connection buys no extra minutes — the clock keeps running and saved answers stay saved. The exam window itself is controlled: the software can require full screen, block copy and paste, stop tab switching, and disable right-click. Access is limited to enrolled students, and each student gets one attempt unless the teacher allows a retake.
Proctoring: live, recorded and AI-assisted
Proctoring is what replaces the invigilator walking the exam hall, and it comes in three shapes. Live proctoring puts a trained human on a video call with the student for the whole exam, watching the webcam feed, the screen and the microphone. Recorded proctoring captures the session for review afterwards. AI proctoring sits in between: software watches the webcam, audio and screen in real time and flags moments — a second face in frame, a voice in the room, the student looking away repeatedly, a tab switch, a paste attempt. The important detail is that a flag is not a verdict. A good system sends flagged moments to a human reviewer who looks at the context before any action is taken, because automated detection gets things wrong — studies of AI text detectors have shown high false-positive rates, especially for students writing in a second language. Most systems also verify identity first: a photo ID check or a face match against the student's registered photo.
Grading: instant for objective questions, queues for the rest
Grading splits cleanly into two paths. Objective questions — multiple choice, true or false, fill in the blank, matching — are scored the moment the student submits. The system compares each answer with the stored key, applies the marks, and can give partial credit on multi-select questions: three out of four correct answers earns three quarters of the marks, not zero. Written answers take a different path. Essays and short answers land in a grading queue where a teacher reads each one, assigns a score against a rubric and adds feedback. Some platforms suggest marks with AI, but the teacher accepts or changes every suggestion — the machine proposes, the human decides. Once graded, those scores join the automatic ones in the gradebook.
Results and analytics: what the numbers actually say
The exam does not end at the score. Per-question analytics show which questions students found hard, which wrong answers were chosen most often, and how long each question took — the evidence a teacher needs to fix a badly worded question or reteach a topic. Results publish in one click: a mark sheet per student, a class rank list, and a department report. Every attempt, flag and grade change is logged with a timestamp, so a disputed result can be traced back to the exact paper and answers.
What good exam software gets right
The best systems are boring in the right ways. They keep one question bank that every exam draws from, so nothing is written twice. They freeze the paper at publish time, so the exam students sat is the exam on record. They save answers continuously and run the timer on the server, so a dropped connection is an inconvenience, not a disaster. They treat proctoring flags as evidence for a human to judge, never as an automatic guilty verdict. And they close the loop, so a teacher learns something from every exam, not just a list of scores.
Frequently asked questions
Can online exams be proctored without a webcam?
Yes. Some platforms monitor browser and behaviour signals instead — tab switches, copy-paste attempts and typing patterns — and produce an integrity score per attempt. It is lighter on privacy and works on any device, though it catches less than a camera-based session.
Does an AI proctoring flag mean the student cheated?
No. A flag means the software saw a deviation from the expected pattern, such as a second face or a background voice. A human reviewer must look at the flagged moment in context before any action is taken, because automated detection produces false positives.
What happens if a student's internet drops during an exam?
Answers are auto-saved as they are entered and the timer runs on the server, so the student can reconnect and continue where they left off. No answers are lost, though the lost time is not refunded unless the teacher allows it.
Do all students get the same questions?
Usually they get the same question set, but in a different order with shuffled answer options — and many exams generate several sets so neighbouring students see different sequences. The paper is frozen at publish time, so every student sits the same recorded version.
Build your online exam platform with InRay Labs
If you run a school, college or training institute and need exam software built around your subjects, your question bank and your rules — proctoring, auto-grading, results and all — InRay Labs builds it. Talk to us about your exam platform and we will scope a system that fits how your exams actually run.