EN
College testing centers are dealing with a different kind of problem than they were three years ago. The questions used to be about scheduling, seat counts and vendor outages. Today the hardest questions are about integrity: students who score perfectly on a recorded exam with nothing suspicious on the video, lockdown browsers that can be bypassed with a tutorial, and smart glasses that look like ordinary eyewear.
The common thread is that cheating has moved from things a proctor can see to things a proctor cannot. Below are the five concerns we hear most often from testing center leaders, and how an AI-assisted approach to proctoring addresses each one.
1. AI-assisted cheating that video review misses
Placement and remote exams are seeing score jumps that are hard to explain. A student who could not break a middle score one week posts a near perfect result a few days later. The recording shows a student sitting quietly at a desk. Nobody is talking to them, nothing is visible on screen, and the answers come without any visible work.
What is happening is simple. The help is coming from a second device, an earpiece or an AI tool that never appears in the camera frame. A human reviewer watching a video has very little to go on, so many institutions respond by pulling remote options altogether. That protects the score, but it costs access for students who cannot easily come to campus.
How it gets solved: Detection has to look at behavior and results together, not just the video.
- Behavior signals. Gaze patterns, reading time, pauses before answers and eye movement toward an off-screen point tell you more than a still frame does. Talview Alvy, built on experience from more than 20 million proctored sessions, flags these patterns in real time.
- Result signals. Score velocity against a prior attempt, answer timing that is too consistent, and correct answers on hard items with no working time are all measurable. Pairing these with session behavior separates a strong student from an assisted one.
- A second camera. A phone used as a side camera covers the desk, hands and surrounding area, which is where most of the help actually sits.
The goal is to keep remote testing available, with evidence strong enough that faculty and administrators trust the result.
2. Lockdown browsers being bypassed
Lockdown browsers were built for a world where the threat was a second browser tab. That world is gone. Short videos showing students how to break out of a lockdown browser are easy to find, and AI overlay tools now sit on top of the screen in ways a blocklist was never designed to catch.
The risk is highest where students test on their own laptops, which is common for accommodated exams. Saved logins, background apps and hotkeys all travel with the device. When one case comes to light, the usual reaction is that an instructor moves the whole course back to paper.
How it gets solved: A secure browser has to watch what the device is doing, not just what it is told to block.
- Behavior over blocklists. The Talview Secure Browser, with Alvy built in, looks for window switching, overlays, unusual key activity and processes that should not be running, even when they are not on any list.
- Live intervention. When something looks wrong, Alvy can pause the session or bring in a human proctor in real time, instead of leaving it for someone to find on a recording days later.
- Safe for personal devices. Accommodated students can keep using their own laptops, with the same level of protection as a center workstation.
This is the difference between a fence and a guard. A fence stops what you planned for. A guard notices what you did not.
3. Smart glasses and wearables in the testing room
Smart glasses have gone from rare to routine at the check-in desk. They carry cameras, microphones and wireless links, and many come with prescription lenses, which makes a simple ban awkward. Centers are rewriting prohibited item policies, asking students to bring a second pair of glasses, and trying phone apps that scan for nearby devices.
Policy helps, but it depends on a staff member spotting the device at the door. Smart glasses are designed to look like regular frames. Smart watches, rings and earbuds are even smaller.
How it gets solved: The testing room itself needs the same AI support that remote sessions already have.
- In-room AI monitoring. Alvy can watch in-person stations through the cameras a center already has, and flag head movement, gaze patterns and gestures linked to wearable use.
- Check-in screening. Wireless device detection at check-in catches active Bluetooth and Wi-Fi devices before the exam starts, instead of after.
- One record. In-person and remote sessions sit in the same review queue, with the same evidence standard, so a case from the testing room is handled exactly like a case from a dorm room.
Staff stay in charge. The AI simply makes sure they are looking at the right station at the right time.
4. Staffing live remote proctoring
Many colleges now run live remote proctoring with their own staff rather than sending every exam to an outside provider. It keeps costs down and keeps the student experience in house. It also stretches teams that were sized for in-person testing only.
The questions directors are working through are practical. How many proctors per session? Do the same people cover the room and the remote queue? What happens at 9 pm during finals week? In most cases, remote testing was added without new headcount, so it comes out of existing capacity.
How it gets solved: Let AI carry the watching, and let people carry the judgment.
- Higher proctor ratios. With Alvy monitoring every session continuously, one proctor can supervise many more test takers, stepping in only when a session is flagged.
- Overflow when you need it. Talview Managed Services can take evening, weekend and peak period sessions, so center staff are not stretched past what they can deliver.
- Works with the tools you have. Sessions run alongside the LMS and exam platforms already in use, so faculty do not have to learn a new workflow.
The result is more remote capacity from the same team, with superior real time support for students when something goes wrong.
5. Check-in and identity at scale
Check-in is where a busy center loses the most time. Staff compare a face to a card, match a name to a booking and log the visit, often with a line out the door. Remote sessions add a harder version of the same problem: confirming the person on camera is the student who registered.
Identity fraud is also changing. Edited IDs and deepfake video are now cheap to make, and a quick visual check by a busy staff member is not built to catch them.
How it gets solved: Make identity a step the system does, with a person ready to step in.
- Automated ID verification. Talview checks the ID document and matches it to a live face capture in seconds, in person or remote.
- Deepfake detection. Live in production, it flags synthetic or replayed video before the exam begins.
- Human escalation. When a match is unclear, a trained reviewer makes the call, so no student is turned away by an algorithm alone.
- Connected check-in. Verification can plug into the scheduling and check-in tools a center already uses, so the result lands in the same record as the booking.
Platform, people and processes
None of these concerns is solved by technology alone. Each one needs three things working together.
- Platform. AI that watches behavior and results in real time, across remote and in-person sessions, and catches more events than anyone else.
- People. Trained proctors and reviewers who step in when the AI flags a session, so every decision that affects a student has a human behind it.
- Processes. Clear policies on devices, evidence and escalation, applied the same way in the testing room and at home.
Testing centers protect the value of every credential their institution awards. The threats will keep changing. With the right platform, people and processes, centers can keep remote and accommodated testing open to students and still stand behind every score.
Want to see how Talview supports testing centers? Talk to our team.
Leave a Reply