Online proctoring for hiring assessments: a practical guide

Online proctoring is the supervision of a test taken remotely, using a webcam, screen monitoring, browser controls or automated analysis to check that the right person completes the test without unauthorized help. In hiring, it helps you trust the scores from remote assessments so you can compare candidates fairly, without asking everyone to travel to a test center.

Proctoring is one integrity control among several, not a complete answer on its own. This guide compares live, recorded and automated proctoring, describes the main integrity risks in remote hiring tests, sets out a seven-step plan to protect your assessments, and covers the privacy, consent and accessibility questions you need to settle before you switch anything on.

What is online proctoring?

Online proctoring, also called remote proctoring, recreates part of what an invigilator does in an exam room. Someone, or something, watches for signs that a test result may not reflect the candidate's own ability: a different person taking the test, notes or a second device on the desk, another person giving answers, or content copied from elsewhere.

In hiring, the context differs from education. Candidates take tests at home, on their own equipment, often outside working hours. The goal is not to catch people out but to make scores comparable, so that a strong result means the same thing for every candidate.

A panel discussion published in Personnel Psychology (Tippins and colleagues, 2006) set out the trade-offs that still apply: unproctored internet testing is convenient and cheaper, but raises concerns about test security, candidate identity and cheating.

Proctored and unproctored testing

It helps to separate three situations:

• Unproctored testing. The candidate takes the test alone, with no supervision. Integrity depends on test design, such as time limits and randomized questions, and on checking results later.
• Proctored testing. The session is supervised in some way, by a person, by recording or by software, and suspicious events are flagged for review.
• Verified testing. Any of the above, followed by a short check later in the process, such as a supervised retest or an interview question that asks the candidate to explain their work.

Most hiring programs combine these. For example, a company might run an unproctored screening test early, then use proctored or verified steps for shortlisted candidates.

Live, recorded and automated proctoring compared

There are three common models. They differ in who watches, when the review happens and how the candidate experiences the test.

Some providers offer hybrid models, such as automated monitoring with a person available to step in. Whatever the model, keep one principle: software flags, people decide. An automated alert should start a review, not end a candidacy.

What proctoring tools can monitor

Depending on the tool and your settings, proctoring can include:

• Identity checks at the start, such as a photo or ID shown to the camera.
• Webcam monitoring for a second person, the candidate leaving the frame, or looking away for long periods.
• Browser controls that limit opening new tabs, copying and pasting, or switching applications.
• Screen activity, such as focus changes away from the test window.
• Audio monitoring for voices.
• Answer-pattern analysis, such as unusually fast correct answers or answers that match known leaked content.

More monitoring is not always better. Each signal collects more personal data and adds friction, so choose the minimum set of controls that addresses your real risks.

Common integrity risks in remote hiring tests

Different tests carry different risks. A speeded reasoning test, a personality questionnaire and a take-home coding task each fail in different ways.

Two points from research are worth knowing. First, not every test is equally easy to cheat on. Arthur, Glaze, Villado and Taylor (2010) studied unproctored internet tests used in selection. For a speeded cognitive ability test, the score differences they found looked more like a normal practice effect than cheating, which the authors linked to the speeded format. Time pressure itself can act as a control.

Second, some risks are not solved by proctoring at all. In the same study, personality scores were higher when the stakes were high, but the pattern was similar to what earlier studies had reported for proctored tests. A webcam does not stop a candidate from presenting themselves favorably on a questionnaire. For self-report measures, item design and how you use the results matter more than supervision.

The growing role of AI tools

General-purpose AI assistants have changed the risk picture for knowledge and coding tests. A candidate can ask a chat tool on a phone for a plausible answer in seconds, and browser controls cannot see that phone. Many organizations respond by favoring applied judgment in test design and adding a short verification step later in the process.

How to protect assessment integrity in 7 steps

The steps below follow the order in which most hiring teams build an assessment process. Adapt them to the stakes of the role and the volume of candidates.

1. Match the controls to the stakes. List your assessment stages and decide how much each score matters. An early screening test that only narrows a long list needs lighter controls than a final test that decides between two finalists. Write down which controls apply at each stage and why.
2. Use large question pools and randomization. When every candidate sees the same fixed questions in the same order, answers spread quickly. Drawing questions from a large, regularly refreshed pool and randomizing their order makes shared answers far less useful. Retire items that show signs of exposure.
3. Set realistic time limits. Timed sections reduce the opportunity to look up answers or consult others, as the research above suggests. Set limits from pilot data, not guesswork, so they test ability rather than reading speed, and adjust them for candidates who need accommodations.
4. Choose a proctoring model and configure it carefully. Pick live, recorded or automated proctoring based on the table above, then switch on only the controls you need. Test the full candidate journey yourself on common devices and connections before launch.
5. Review flags with people, not rules. Decide in advance who reviews flagged sessions, what evidence they look at and what counts as a confirmed issue. Record the decision and the reason. Treat a flag as a question to answer, not as proof, and offer a retake when the cause is technical.
6. Verify in the interview. Ask candidates to talk through one or two of their answers, extend a coding task or solve a related problem live. Someone who did the work can usually explain it. This step also reduces the impact of any controls that did not work, and fits naturally into a structured, skills-based hiring process.
7. Tell candidates what to expect. Explain before the test what will be monitored, why, who sees the data and how long it is kept. Share technical requirements, a practice test if possible, and a contact for problems. Clear rules reduce both honest mistakes and anxiety.

Choosing a platform with integrity features

When you compare tools, ask how each one handles question pools, randomization, time limits, proctoring options and the review of flagged sessions, not only whether it "has proctoring". The guide to pre-employment assessment tools covers the wider selection criteria.

For example, TeamCheck's Technical Assessments include automated anti-cheating checks and AI proctoring analytics that flag non-authentic attempts, drawing on a library of more than 115,000 questions across 25,000+ ready tests, so results stay fair and comparable.