What does Human-First proctoring actually mean?

Human-First means that a real proctor is present in real time during every exam. Certory uses no automated AI systems for behavioural analysis – all observations and assessments are made by a human being who is live in the session.

Why doesn't Certory use AI-proctoring?

We are convinced that human judgement in high-stakes exam situations creates more accountability and fairness. A person can read context, recognise exceptions, and respond when needed – in ways that automated systems simply cannot replicate.

How quickly is a proctor available?

On average, candidates are connected to a proctor in under five minutes. In practice, it is usually even faster.

Is human proctoring more reliable than automated proctoring?

What human live proctoring offers above all is accountability: a proctor documents what they have observed. That makes the process explainable – for the institution, for the candidate, and for anyone who needs to understand the result.

What kinds of exams is Certory designed for?

Certory supports high-stakes online assessments of all kinds: final exams at universities, professional certification tests, corporate compliance assessments, and licensing exams. All that is needed is a standard browser. No plugin, no IT project.

Certory ThoughtS - Welcome to the human-first proctoring blog

WHEN IT COUNTS

An alibi "human in the loop" isn't enough - not for the exams that decide a future!

AI is extraordinary. It writes, codes, diagnoses, translates and designs - and we're as excited about it as anyone. But not every job should go to a machine just because it technically can.

Supervising an exam that decides someone's future is one of those jobs. And recently, the world got a very public reminder of what happens when that line gets blurred.

Because when it really counts, it matters who is there - and how many of them.

Get in contact

A young woman with long brown hair makes a pouting facial expression against a bright blue background.

AI Proctoring + one supervisor for every 150 candidates?

Earlier this year, a large public university in Mexico ran its admissions exam fully online for the first time. Tens of thousands of candidates, supervised by AI: a locked-down browser and AI-webcam monitoring, with software flagging anything it found unusual.

Then the results came in. Top scores that used to be rare suddenly appeared at several times the historical rate. The pattern was hard to explain — and impossible to ignore. In the end, tens of thousands of candidates were asked to sit the exam again, this time in person, with people in the room. The start of the semester had to be pushed back to make it happen.

One detail stood out: on paper, there was a "human element": one supervisor was assigned for every 150 applicants, notified by the system whenever it flagged something. On paper, there was a "human in the loop."

In practice, one person cannot meaningfully watch 150 live exams at once. That isn't supervision. That's an alibi.

"AI is an effective cheater, not a supervisor"

A quote form the article brings out the uncomfortable irony: the same kind of technology that's meant to catch cheating is remarkably good at enabling it. AI can draft an answer in seconds, quietly, off-screen, in a second tab, on a second device.

Detecting that reliably, in real time, across thousands of simultaneous sessions, is a completely different problem from generating it. An algorithm can flag a glance to the side or an unusual noise. What it cannot do is understand suhc a vast human context, exercise sufficent judgement, or step in with a human touch at the right moment.

For the moments that decide someone's future, that gap isn't a technical detail. It's just does not live up to the requirements!

Book your Demo

get in touch with us today

What a "human in the loop" quietly leaves out

"Human in the loop" sounds reassuring. It suggests a person is watching, deciding, accountable. But the phrase can quietly hide the two things that actually matter: presence and ratio.

Being notified by a system is not the same as being there. A supervisor who is pinged after the fact, or stretched across 150 simultaneous exams, isn't supervising in any real sense — they're rubber-stamping whatever the software happened to surface.

Real human supervision isn't a checkbox at the edge of an automated system. It's the system.

What real human supervision looks like

At Certory, a real proctor is present for the entire exam — live, in real time, from start to finish. Not one for 150. Our average ratio is roughly one proctor to six candidates.

It starts with a calm onboarding: standard browser, no plugin, connected to a proctor within minutes, no AI, no biometrics. During the exam, two cameras — PC and mobile — give a complete view. The proctor stays in the background: no pop-ups, no automated suspicion flags, no interruptions. Everything relevant is documented with a timestamp, and if anyone later asks what happened during that exam, there is an answer, added by someone who was actually there.

The result of designing around people instead of suspicion: 86% of our candidates rate their experience as positive, and fewer than 5% need support during the exam. Calm, focused, fair and genuinely supervised.

This isn't about one institution getting it wrong. It's about a model that treats human oversight as a formality to point to, rather than as the thing that carries the weight. Because a "human in the loop" sometimes just isn't enough.

For the moments that decide someone's future, trust shouldn't be automated. It should be fully human.

Frequently asked questions

You ask, we answer.

What does "human in the loop" mean in online proctoring?

It describes a setup where AI monitors the exam and a human is involved only at the edges — reviewing flags afterwards, or overseeing many exams at once. The person is present in name, but not necessarily watching your specific exam as it happens. It is very different from live human supervision, where a proctor is present throughout the session in real time.

Is a "human in the loop" enough for high-stakes exams?

A supervisor responsible for 150 simultaneous exams cannot meaningfully observe any single one of them. For high-stakes exams, where the result carries real consequences, continuous human presence at a workable ratio makes a decisive difference to fairness and accountability.

Why did so many candidates recently have to resit an online exam?

In a widely reported case, a fully online, AI-supervised admissions exam produced score patterns so far outside the historical norm that the institution asked a large number of candidates to sit the exam again in person. The automated setup — including a very high candidate-to-supervisor ratio — did not hold up.

What proctor-to-candidate ratio does Certory use?

On average, roughly one proctor to six candidates. A live proctor is present for the entire exam in real time — not notified after the fact, and not spread across hundreds of simultaneous sessions.

Should AI supervise a high-stakes exam on its own?

AI can support a process. But reliably supervising in real time and understanding such a broad context as a human in an exam situation is a very different problem. For exams where the outcome truly matters, we believe a real person should be in charge.