Remote examinations were supposed to democratise access to opportunity. No travel, no test centres, no geographic disadvantage. In theory, a student in a small town could sit the same examination as someone in a major city and be evaluated on equal terms.

In practice, a thriving industry has emerged to defeat that promise.

The new economics of cheating

The latest generation of AI-powered exam-assistance tools does not require a hidden earpiece, a second phone under the desk, or a friend whispering answers through a window. It requires a laptop, a subscription, and thirty seconds of setup.

The tool listens to the examination through the microphone, sends the question to a cloud AI service, receives an answer, and displays it as an overlay on the candidate's screen — positioned where the eyes naturally fall, refreshing silently as new questions appear. The barrier to entry has collapsed. What once required coordination, risk, and a co-conspirator now requires a credit card.

Invisible by design, not by accident

The overlay is not invisible in the way that something small is hard to see. It is invisible in the technical sense: the tool instructs the operating system's display compositor to exclude its window from all screen capture. The screen-sharing software the proctor is watching receives a frame with a blank space where the AI answers sit.

The proctor sees a calm candidate thinking carefully. The candidate is reading.

This is the crucial point that the examination industry has been slow to confront. The cheating is not happening in a blind spot the proctor could catch with more attention. It is happening in a region of the screen that has been surgically removed from the image before the proctor ever receives it.

An open, advertised industry

This is not a fringe capability traded in dark corners. These tools are marketed openly, reviewed on video platforms, and used in job interviews, professional certifications, and academic examinations worldwide. Their developers publish comparison charts showing exactly which examination platforms they are invisible on — treating the proctoring industry's blind spot as a product feature.

When a category of fraud is this accessible, this affordable, and this openly marketed, it stops being an edge case. It becomes the baseline assumption every high-stakes examination has to plan around.

Why current proctoring cannot respond

Current remote proctoring operates by recording or monitoring a screen-share feed. That feed is produced by the same screen-capture interfaces that these tools are designed to evade. The proctor is watching an edited version of reality — one from which the cheating has already been removed before the image is transmitted.

No amount of AI analysis applied to that feed can recover what was never in it. You cannot detect, in a recording, content that was excluded before the recording was made. Better monitoring of the same feed is not a solution. It is a more expensive way to watch the wrong thing.

What a real solution requires

The answer is not better monitoring of the same feed. It is capturing the feed before the editing happens — owning the capture layer rather than borrowing it from a video-conferencing platform. It requires building examination-delivery infrastructure from the ground up with integrity as the primary requirement, not an afterthought bolted onto a video call.

That is the principle behind OroLink, our open-source secure examination-delivery protocol: capture the complete frame at a stage of the pipeline that precedes the point where these tools hide, and show the proctor what the candidate actually sees.

Key takeaways

  • Modern AI exam-assistance tools render answers as overlays that are deliberately excluded from screen capture.
  • Any proctoring product built on screen sharing receives a frame with the cheating already removed.
  • The fix is architectural — capture before the exclusion happens, not smarter analysis of an already-edited feed.

The examination is not over. The arms race has just begun.