How PassPilot works, in detail.

Every PassPilot paper starts with the official exam pattern guide. We read it cover to cover, mapping out section durations, question types, difficulty distribution, and scoring rules. If a section has 40 questions in 60 minutes, ours does too.

Then we collect real past papers where the exam body releases them — IELTS publishes a few, OPRA publishes a few, GRE publishes many. We catalogue every question type we find and rebuild a question bank organised by topic and difficulty.

Each new paper is reviewed by an exam specialist — someone who has either taken the exam themselves or has 1000+ hours of teaching experience. If a question doesn't feel right, it gets cut. We typically reject 1 in 5 questions in review.

Finally, the paper is pilot-tested by 30–50 users. We look at completion time, score distribution, and user feedback. If a section takes 20% longer than the official time, we cut the hardest 2 questions and replace them.

When you click "Start test", the server loads a fresh question set from our bank. You cannot get the same question order as another user taking the same paper — every attempt is randomised.

The timer runs on the server, not in your browser. That means refreshing the page, switching tabs, or losing internet doesn't give you extra time. When the timer hits zero, the test is auto-submitted.

Your answers are auto-saved every 30 seconds. If you lose connection, you reload, return to the same question, and the timer continues from where it stopped. No work is lost.

For audio questions, your microphone is requested when you first hit that question. We record your answer, upload it to our servers, and process it for scoring. Nothing is recorded without your explicit permission.

For objective questions (MCQ, true/false, matching), scoring is deterministic. We compare your selected answer against the correct one. If negative marking is enabled for that test, we deduct points for wrong answers.

For subjective questions (essays, short answers, writing tasks), we use a multi-step automated assessment. First, your response is evaluated against a rubric — typically 4 criteria with weights. Second, we generate specific feedback on what you did well and what to improve. Third, the assessment is verified for consistency with what a human examiner would give. Results land in under 60 seconds.

For audio questions (speaking tests, OET role-plays, TOEFL integrated tasks), we first transcribe your speech to text. Then we evaluate the transcript the same way we evaluate a written answer. We also analyse fluency markers, vocabulary range, and grammar complexity. The whole thing is processed live, so you see your band score before the timer for the next section starts.

Every score is logged with a confidence indicator. If our system is uncertain (e.g., a borderline answer), we flag it and a human reviewer may re-score it within 24 hours.