What candidates see

The consent and preflight gate

Before starting, every candidate passes a consent gate that states exactly what is monitored during the attempt — for example webcam or screen recording when your delivery enables them. When the delivery uses camera proctoring, screen recording, or ID verification, a preflight check also verifies the candidate's browser, connection, and camera. Candidates only enter the exam once the gate is cleared, so technical problems surface before the clock starts rather than mid-attempt.

The exam view

The exam view keeps orientation cues visible at all times. A sticky top bar shows the assessment title, an autosave indicator, and TIME LEFT with a progress bar — it turns amber in the final two minutes and red in the final thirty seconds. A question navigator summarizes progress: short assessments show a grid, while multi-section or longer assessments show a collapsible section tree. Either way, the current question is highlighted, answered questions are marked, a corner bookmark indicates a question marked for review, and hovering shows the question title.

Candidates move through the assessment page by page using Prev and Next buttons and page pills. When the open page has its own time limit, a PAGE timer chip appears so the candidate can pace that page separately from the overall countdown.

Marking questions and reviewing before submit

Each question carries a mark-for-review flag. Flagged questions show a corner bookmark in the navigator, making it easy to return to them later. Before submitting, candidates see a review dialog with answered, unanswered, and marked counts, plus a chip for every unanswered question; in free navigation each chip jumps straight back to that question before committing.

Autosave

Responses autosave as candidates work. There is no manual save step, so a refresh or brief interruption does not discard in-progress answers. Note that shuffle order is seeded per attempt: if your assessment shuffles sections, pages, or blocks, the same candidate sees the same order after a refresh, while different candidates see different orders.

Voice interview tasks

In a voice interview task, the candidate speaks for up to about three minutes per turn. A transcript is stored as evidence, and an adaptive follow-up question is generated and spoken back with subtitles. A typing fallback is always available for candidates who cannot or prefer not to speak.

The interviewer is a disclosed AI persona, labeled "AI interviewer", with a talking-avatar clip for the opening question. In evaluation, only the candidate's own words count as evidence — accent and fluency are excluded by design.