How online proctoring works: technology, process, modes

Online proctoring allows universities, certification bodies, employers, and public-sector organisations to supervise exams when candidates are not in the same room as an invigilator. It combines identity checks, system controls, webcam and microphone monitoring, desktop activity, and human or automated review. The challenge for institutions is choosing a process that protects assessment credibility without creating unnecessary friction, support work, or another disconnected tool.

Key takeaways

• Online proctoring combines identity verification, system checks, monitoring, event detection, and session review
• Candidates usually complete a browser, device, camera, microphone, internet, and identity check before starting
• Monitoring can combine webcam video, microphone audio, desktop activity, secure-browser controls, and device signals
• Institutions can choose AI-only, post-exam review, live review, or identification-focused proctoring based on exam risk
• A flagged event is not always proof of a violation, so clear review and escalation processes remain important

• Buyers should evaluate integration, privacy, accessibility, scalability, support, and the complete candidate experience

   

What is online proctoring?

Online proctoring is the supervision of a remote exam using technology instead of, or alongside, an invigilator in the same physical room. The system may verify identity, monitor the candidate through a webcam and microphone, track activity on the device, and record events for automated or human review. It is used by universities, schools, certification bodies, employers, ministries, and government training programmes where assessment results need consistent oversight.

The level of monitoring should reflect what is at stake. A short employee knowledge check may need lighter controls than a final university exam, professional licence, or regulated certification. Wider developments in AI in higher education have also increased interest in using automated monitoring to support large and distributed candidate groups.

How online proctoring works, step by step

The details vary between providers, but most remote proctoring systems follow a similar journey from candidate login to final reporting. Understanding how online proctoring works helps assessment and IT teams identify where technical checks, human decisions, integrations, and support responsibilities sit. The process should be tested as a complete workflow rather than treated as a single monitoring feature.

Getting started: browser-based access, no installs, LMS integration

Many online proctored exams start through a browser link or directly from an institution’s LMS. Candidates sign in, review the rules, provide the required permissions, and begin the onboarding process without installing a full desktop application. Some configurations may still use an extension, secure-browser controls, or a mobile component, so institutions should explain the exact requirements before exam day.

LMS integration can connect candidate enrolment, exam access, authentication, results, and review workflows. This reduces the need for staff to maintain separate candidate lists or ask learners to create additional accounts. Buyers should still test how data moves between systems, how errors are handled, and what happens when the LMS or proctoring connection is temporarily unavailable.

System checks: camera, microphone, internet, second-screen detection

Before the exam opens, the system checks whether the candidate’s camera, microphone, browser, and internet connection meet the required conditions. It may also identify extra displays, prohibited applications, virtual environments, or screen-sharing tools, depending on the platform and exam rules. Candidates who fail a check should receive clear instructions for resolving the problem rather than reaching a generic error screen.

A practice check before the assessment window can reduce avoidable disruption. It gives candidates time to update permissions, change devices, test their connection, or contact support. Institutions should include accommodation workflows in the same testing process so approved assistive technology is not incorrectly blocked.

Identity verification: photo ID and face match

Identity verification helps confirm that the registered candidate is the person beginning the exam. Depending on the policy, this may involve photographing an approved identity document, capturing a live image, completing automated face matching, or speaking with a human proctor. Only the minimum identity data needed for the assessment should be collected and retained.

Different candidate groups may need different verification routes. A university may accept a student card, while a certification body may require a government-issued document. Institutions should also provide a fallback process for unclear images, expired documents, name differences, accessibility needs, or candidates who cannot complete automated matching.

During the exam: combined data from camera, microphone, and desktop

Once the assessment starts, the platform may combine webcam video, microphone audio, desktop activity, browser events, and device information. This can help identify events such as another person appearing, the candidate leaving the frame, unexpected speech, a tab change, a blocked application, or an additional device. The exact signals monitored should be stated clearly in the candidate instructions.

Monitoring does not always mean that a person watches every candidate continuously. In an automated model, software records and flags events for scoring or later review. In a live model, AI alerts can help a human proctor direct attention towards sessions that may need immediate support or intervention.

After the exam: consolidated session record and review

After submission, the system brings together the available camera, audio, desktop, identity, and event data into a session record. Reviewers may see timestamps, recordings, event types, and risk indicators, allowing them to examine relevant moments rather than replaying every session from beginning to end. The final workflow may be automated, reviewed by a person, or use both approaches.

Institutions should define what happens after a flag appears. Technical issues, background interruptions, approved accommodations, and potential rule violations should not automatically receive the same response. Clear review criteria, escalation routes, evidence requirements, and appeals help make decisions more consistent.

The online proctoring technology behind the scenes

Online proctoring technology brings together several components rather than relying on one detection model. These may include computer vision, voice detection, face matching, browser controls, device checks, screen recording, event logs, and risk scoring. Readers looking specifically at what AI proctoring is should distinguish between the AI that identifies signals and the institutional process that decides what those signals mean.

For example, the system may detect that a face disappeared, another voice was present, or the exam window lost focus. It cannot always determine why that happened or whether the event represents misconduct, a technical problem, or ordinary behavior. Strong review tools therefore connect each flag to a timestamp and supporting evidence so an administrator or proctor can examine the event in context.

Some platforms use an overall risk level to help reviewers prioritize sessions. This can reduce the time spent opening low-risk recordings, but it should not replace a documented review process for consequential decisions. Institutions should understand how the score is calculated, whether thresholds can be configured, and how false positives are identified.

What online proctoring flags as a violation

What counts as a violation depends on the exam policy. A calculator may be permitted in one assessment and prohibited in another, while speaking aloud may be acceptable for an accommodation but flagged under a standard rule set. The platform should allow institutions to configure rules by assessment rather than applying one inflexible policy everywhere.

Common automated signals include:

• The candidate’s face disappearing from the camera view
• An additional person or face appearing
• A mismatch between the candidate and registered identity
• Repeated or unexpected voice activity
• Looking away from the screen beyond a configured threshold
• Opening another browser tab, window, or application
• Disconnecting or adding a monitor or device
• Attempting to use screen-sharing, remote-access, or virtual-machine tools

Human proctors may also identify physical notes, prohibited calculators, phones outside the main camera view, non-verbal communication, or behavior that automated models cannot interpret reliably. These events should be recorded with timestamps and reviewed against the published rules. A flag should direct attention to an event, not automatically determine the outcome.

The four modes of online proctoring

Institutions do not need to use the same level of supervision for every exam. Most platforms offer different modes that balance scale, cost, flexibility, and human involvement. The right choice depends on the consequence of the result, candidate volume, available reviewers, and whether immediate intervention is required.

• Automatic or AI-only proctoring: The system monitors the session and produces results without a human proctor reviewing every exam. It can support flexible scheduling and large candidate volumes, but it provides less contextual judgement.
• Asynchronous or post-exam review: AI monitors and records the assessment, then a human reviewer examines flagged events after submission. This adds judgement without requiring candidates and proctors to attend at the same time.
• Synchronous or live proctoring: A human proctor supervises candidates in real time, often supported by AI alerts. This is useful where immediate intervention, communication, or manual checks are important.
• Identification-only proctoring: A human confirms the candidate’s identity before the assessment, while the remaining session is monitored automatically. This can suit exams where identity assurance matters but continuous live supervision is not required.

AI-only monitoring can be appropriate for lower-risk or high-volume assessments, while human review is often easier to justify for high-stakes decisions. A hybrid approach can also help teams use automation for broad coverage while retaining people for uncertain or consequential cases.

Computer vs. phone: what devices work for online proctoring?

Desktop and laptop computers usually provide the strongest environment for an online proctored exam because they support full-screen assessment delivery, browser controls, desktop monitoring, and more stable camera positioning. Mobile devices can be useful where candidates do not have computer access or where the platform uses a phone as a second camera, but the available controls may differ. Constructor Proctor also offers mobile capabilities for supported post-exam recording workflows, so buyers should confirm which devices work with each proctoring mode rather than assuming every feature is available everywhere.

Device policy also affects accessibility and participation. Requiring a modern computer, strong internet connection, specific operating system, or second camera may exclude some candidates unless alternatives are available. Institutions should publish minimum requirements early and provide practice checks, loan devices, testing centers, or alternative arrangements where appropriate.

How to choose online proctoring technology

Choosing online proctoring technology requires more than comparing lists of AI detections. Your team should evaluate the candidate journey, administrator workload, review process, LMS integration, accessibility, data handling, and performance during peak exam periods. The right online proctoring software should reduce operational complexity rather than create another isolated system.

Ask the vendor to demonstrate a complete exam rather than only the administrator dashboard. A useful pilot should include different devices, internet conditions, accommodations, late starts, identity failures, disputed flags, and peak candidate volumes. Your implementation plan should also state what the vendor manages and what remains with internal assessment, IT, privacy, and student-support teams.

How Constructor Proctor fits in

Constructor Proctor supports AI review, post-exam human review, and live proctoring within a platform built for education, certification, and workforce assessment. It supports more than 10,000 simultaneous exam sessions, over 30 LMS integrations, a live proctor-to-candidate ratio of up to 1:150, and monitoring across more than 100 behavioral and technical parameters. It also includes secure-browser controls, device detection, identity verification, special-accommodation settings, and APIs for connecting proctoring with existing assessment environments.

As part of Constructor Tech’s all-in-one platform, Proctor can connect with learning, assessment, data, and reporting workflows rather than operating as a separate exam-security tool. This integrated ecosystem is intended to help institutions reduce fragmented tools while keeping their teams in control of review decisions. At Princess Nourah University’s English Language Institute, a wider digital assessment implementation that included Constructor Proctor reported an 85% increase in student engagement, a 70% reduction in administrative workload, and a 90% drop in reported academic dishonesty; those results reflect that specific implementation rather than a guaranteed outcome for every institution.