“Was it really them?”
The candidate confirms their email at login and again before the attempt. A reference photo is checked for a clear face, and the face on camera is compared with it during the exam.
Identity checks
AI exam proctoring
Conduct assessments you can trust.
Integrated with your ADFS, LMS or any other HR system, Procterify checks every candidate before the exam, monitors the session with Procterify AI, and scores each attempt with evidence your team can review.
Before the exam
Start stays locked until camera, face and mic pass
During
Answering pauses if the camera can't see the candidate
Evidence
A 15-second clip with audio for significant events
After
A Trust Score and risk level on every attempt
The hard part
Running the exam is easy. Defending the result is harder.
When a result is challenged, someone has to answer three questions. Procterify is built so you can.
“What actually happened?”
Every flag has a timestamp and a photo. Significant events also get a 15-second clip with audio, so you see the moment itself, not just a label.
Evidence
“Was the process fair?”
A dim room or a quick glance away is logged, not punished. Answering only pauses when the camera can't see the candidate, and the screen tells them why.
Candidate rules
How it works
What happens before, during and after every attempt.
Candidates know the rules before they start. Your team gets a clear record after they finish.
Before the exam
- 01
Confirm the account.
The candidate sees the email they signed in with and confirms it, or signs out. They confirm again right before each attempt.
- 02
Take a reference photo.
The photo is only accepted when the face is visible and clear. Procterify creates a facial reference from it automatically.
- 03
Pass the readiness check.
Camera on, face visible, and a real microphone test: the candidate speaks and Procterify confirms it can hear them. Start stays locked until all three pass. If one fails, they fix it and try again on the same page.
During the exam
- 04
AI watches the session.
It looks for a missing or partly hidden face, more than one person, looking away, a blocked or blurred camera, low light, and phones or laptops in view. The face on camera is compared with the reference photo at intervals.
- 05
Serious problems pause answering.
If the camera is blocked, too blurred, can't find a face, or sees more than one, answering pauses and the screen says why. It resumes once the problem is fixed.
- 06
One watermarked session.
A faint watermark with the candidate ID and time covers the exam screen. If the exam is finished in one tab, any other open tab is closed out and can't send more evidence.
After the exam
- 07
Every attempt gets a Trust Score.
A score out of 100 and a risk level, worked out from what was detected, how often and for how long.
- 08
Your team reviews the evidence.
Open any attempt to see each event with its photo and clip, in order.
- 09
Decisions are on record.
Tag candidates for follow-up. The log shows who added each tag and who removed it.
Trust Score
A score that shows its working.
Every completed attempt gets a score out of 100 and a risk level. The score comes from what was detected, how often, and for how long, so any number can be traced back to the events behind it.
Factor 01
Severity
Some events matter more than others. A second person on camera counts for more than a dim room.
Factor 02
Frequency
Repeats raise the penalty, but the effect levels off, so one flickering detection can't run away with the score.
Factor 03
Duration
A two-second glance counts for less than half a minute away from the screen.
0–50 · High risk
51–79 · Medium risk
80–100 · Low risk
Some events go straight to High risk.
A second person on camera, a phone in view, a camera blocked for a long stretch, a face missing for a long stretch, or a virtual camera caps the score at 50, whatever else happened. The score before the cap is kept, so reviewers can see exactly what changed.
Trust Score
Medium risk76
Score 76, Medium risk.
Looking away3 × 3.94 × 2.0 = 23.7
Score100 − 23.7 = 76
The Trust Score sorts attempts for review. It doesn't decide misconduct. A person does.
Evidence
Review the flagged moments, not the whole recording.
Open an attempt and see each event in order, with its photo and, for significant events, a short clip. Repeats of the same ongoing incident are grouped, so you aren't scrolling through duplicates.
- A photo for every event, stamped with the time it happened.
- 15-second clips with audio for significant events, when the microphone is available.
- Identity checks listed separately, so a face mismatch never hides inside an average.
- Tags with a history. Mark candidates for follow-up; the log shows who tagged and who removed it.
Candidate experience
Strict where it matters. Calm everywhere else.
Genuine candidates shouldn't lose an exam to a dim room or a slow microphone. Procterify only stops answering when the camera can't do its job. Everything else is recorded for your team to judge.
Pauses answering until fixed
- Camera blocked
- Camera image too blurred
- No face detected
- More than one face
Logged for review, exam continues
- Looking away
- Low light
- Phone or laptop in view
- Face doesn't match the reference photo
- Microphone stops working
- Clear reasons on screen.When answering pauses, the candidate sees why and what to do.
- Fewer false alarms.Extra checks run when a second face appears, so a poster or photo on the wall is less likely to raise an alert.
- No lost exams to a mic.If the microphone drops out, the candidate is told, the exam carries on, and the gap is recorded.
These rules are agreed with your team during rollout.
Assessments
Build the exam where you proctor it.
The question bank and assessment setup are part of Procterify, so there's no second system to connect and no export to get wrong.
MCQ and passage questions.
Link several questions to one passage. Candidates and reviewers see the passage next to each question.
Set the difficulty mix.
Choose the share of easy, medium and difficult questions for each assessment, for example 45 / 30 / 25.
Sections by category.
Organise questions into categories and draw from each one in the order you set.
Spot gaps before exam day.
A summary shows questions available, configured and actually shown, by category and difficulty. Shortfalls are highlighted.
Short on questions? It still builds.
When a category runs short, Procterify can fill from neighbouring categories or difficulty levels.
Export to Excel.
Download the question bank, long passages included.
Adaptive testing
The next question depends on the last answer.
Give Procterify a question bank and it picks each question as the exam goes. Candidates who keep getting it right move on to harder questions. Candidates who keep getting it wrong get easier ones.
Right answers step it up.
Answer correctly and the next question comes from a harder level.
Wrong answers step it down.
Miss one and the next question is easier, so the result still shows what the candidate does know.
One bank for everyone.
Every candidate draws from the same questions you set up, tagged easy, medium or difficult.
Any question format.
Text, media and data questions all run in the same proctored session, with the same checks and the same Trust Score.
Multiple choice
One or more correct answers.
Passage-based
Several questions on one passage.
Video
Watch a clip, then answer.
Audio
Listening questions with a built-in player.
Image
Diagrams, photos and scanned figures.
Table-based
Read and work from a data table.
Graph and chart
Interpret a plotted chart or trend.
Written answer
Short or long text responses.
Working with a format that isn't listed here? Bring an example to the demo.
Security
Built to deter leaks and close loose ends.
Protections that are honest about what a browser can and can't do.
Traceable screens.
A faint watermark with the candidate ID and time covers the exam, so any photo or screenshot points back to who took it and when.
Screenshot detection.
Common screenshot shortcuts are detected where the browser allows. No browser can catch every method, so we don't claim to; the watermark covers the rest.
One live session.
Finish the exam in one tab and every other tab stops recording and is sent away. The server refuses evidence from an ended attempt.
Virtual camera detection.
Software that fakes a webcam feed is detected and flagged as a critical event.
An audit trail for admins.
See who created each module and test, and who tagged or untagged each candidate.
Retention you decide.
How long clips, photos and facial reference data are kept is agreed with you before rollout.
Candidates see what is checked before they start. Evidence is used to review that attempt.
Use cases
For teams whose results get questioned.
Employee assessments
- Mid-year and annual assessments
- Internal certification and promotion tests
- Category sections and difficulty mixes per role
Pre-employment tests
- Candidates test from home, on their own laptop
- The same checks for every applicant
- A score to sort who needs a closer look
Certification and licensing bodies
- Remote credential exams
- Evidence ready for reviews and appeals
- Decisions recorded with who made them
Universities and training institutes
- Semester, entrance and placement exams
- Passage-based and MCQ papers from one bank
- Fair rules for students on home connections
FAQ
Questions, answered.
No. Procterify records events, attaches evidence and scores each attempt so your team knows where to look. A person reviews the evidence and makes the decision.
Only problems that stop the camera from seeing them: a blocked camera, a very blurred image, no face in view, or more than one face. Answering resumes once it's fixed. Looking away, low light, objects in view and microphone problems are recorded but don't stop the exam.
Each type of event has a severity. The penalty grows with how often it happened and how long it lasted, then is subtracted from 100. A score of 80–100 is low risk, 51–79 medium and 50 or below high.
More than one face for 10 seconds or more, a phone in view for 10 seconds or more, the camera blocked for 20 seconds or more, no face for 30 seconds or more, or a virtual camera. The score is capped at 50 and the original score is kept for reviewers.
The candidate confirms their email at login and before each attempt, takes a reference photo that must show a clear face, and is compared with that photo at intervals during the exam.
Without a working camera and visible face, the exam can't start; the candidate can fix it and retry on the same page. If the microphone stops during the exam, they're alerted and can keep going. The gap is recorded and clips from that time may have no audio.
Procterify detects common screenshot shortcuts where the browser allows, but no web app can block every method. Every screen carries a watermark with the candidate ID and time, so any capture can be traced.
Procterify runs in a modern web browser with a webcam and microphone.
You decide. Retention for clips, photos and facial reference data is agreed with your team before rollout.
Yes. Build your question bank with MCQ and passage-based questions, set the difficulty mix per assessment, and export the bank to Excel.
Book a demo
See Procterify in action.
A live walkthrough with our team. Bring a sample of how you assess today and we'll show you the same exam in Procterify.