Who reviews your exams? Inside the human review behind every result

When a result gets challenged, the first question is rarely about the algorithm. It is about people: who looked at this session, what they were trained to look for, and how you know their judgment was sound. For a director of certification or a registrar, those questions can arrive months after the exam, from a regulator, an appeals panel or the candidate, and the answer needs to already be on record.

Much of the online proctoring market now says that humans make the call. Far fewer explain who those humans are. That gap matters, because human review only makes a result defensible when the review itself can be described, checked and repeated. Here is how review works at Integrity Advocate, from the people who do it to what happens when a decision is questioned.

Our reviewers are our own team

Every session is reviewed by Integrity Advocate’s own full-time reviewers. They are not an agency or a pool of contractors, and they are not your staff working through a queue at the end of the day. A team that trains together, works to the same standards and answers to the same leadership produces judgments you can compare from one session to the next.

All reviewer work runs through Integrity Advocate’s Canadian infrastructure. Access requires a named account with multi-factor authentication and is limited to the sessions each reviewer is assigned, and everyone with access has signed an NDA and passed a criminal background check.

Every session gets a full review

Automated detection has a useful job: it points a reviewer to the exact moment something in a session looked different from expected. A detection is a prompt rather than a verdict, though, so at Integrity Advocate every session is reviewed in its entirety by a person who checks each detection against the actual frames, with the full context of the session around it.

Context is what turns a signal into a finding or rules it out. A reviewer can recognize that a candidate looked away to think, that a young child wandered through the room, or that an item on the desk was one the program allows, where a detector would only register movement or an object. Each review ends in a written outcome recording what was flagged, what the reviewer observed and the reasoning behind it, and there is no automated pass or fail.

How reviewers are trained

No one reviews a live session until they have completed structured onboarding. After that, reviewers take part in ongoing refreshers, which keep standards current as new tools and new behaviours appear, including AI-assisted methods that leave little visible trace.

This is also where human review earns its place as AI tools spread. Because a trained person examines each session, a new or uncommon tool can be recognized the first time it appears, instead of waiting until an automated system has seen it often enough to learn it. Once reviewers understand it, it is added to detection.

A second reviewer checks the work

The majority of sessions are audited by a second reviewer, who confirms the details of the original review. That audit does two things at once. It confirms results before anyone relies on them, and it keeps the whole team consistent, because any difference in judgment surfaces quickly and feeds straight back into training.

How serious cases are escalated

Some situations deserve more than one person's view before they reach a program. When a review raises a serious concern, it moves from the reviewer to a team lead and then to the client, with the evidence attached. The program receives a documented case it can follow from start to finish, rather than an alert it has to investigate from scratch.

The program makes the decision

Integrity Advocate reviews each session against the rules a program sets, at the enforcement level the program chooses, and the final academic or certification decision always belongs to the program. IA never makes that call. The reviewer's job is to give the people who own the decision clear, fair evidence they can stand behind.

Test takers see the same evidence you do

A defensible result also depends on what the test taker can see. By default, each participant receives an email with their session status, and if anything was flagged, that email includes the flagged screens and the name of the rule that was triggered. The participant sees the same proof the program sees, and programs can turn this notification on or off.

If a participant disagrees with a flag, they raise it with the program, which owns the decision. Because both sides are working from the same record, that conversation starts from evidence instead of suspicion.

Thorough review without the wait

Turnaround is set by each customer. The standard for non-urgent work is 24 hours, and in practice results usually arrive within hours. If a program needs results faster, we agree on a turnaround up front. For large exams that run on a single day, we plan extra reviewer shifts with the client in advance. Either way, every session still gets a full human review.

Privacy is part of the review

More people reviewing sessions can sound like more privacy risk. At Integrity Advocate, reviewers are part of the protection, because they blur and block categories of incidental personal information before a session reaches the client's portal, so programs see what they need to decide and nothing more. Data is encrypted in transit and at rest, we are SOC 2 Type II certified, and unflagged sessions are deleted after 24 hours by default. The full detail is in the Security & Privacy Overview.

When a result is challenged

Standards bodies are asking the same questions buyers are. ISO/IEC 17024:2026 asks certification bodies to show documented human oversight wherever AI influences candidate monitoring, with an accreditation clock already running.

So picture that challenge arriving again. With this model, your answer is that a trained member of our own team reviewed the whole session against your rules, a second reviewer checked the work, any serious concern reached you with its evidence, and the test taker saw the same record you did. That is a result you can explain, and one that holds up as fair, trustworthy and defensible.

{{post-cta}}