Compare
Academic integrity tools come in three shapes: score the finished text, lock down the exam environment, or read how the work was produced. They fail in different ways, and the differences decide what your students experience and what your reviewers can actually defend.
01 · Three approaches
After the fact
Paste in finished writing and get back a score for how much of it "sounds" like AI. The judgment is statistical: there is no record of how the text came to be, only a guess about its style.
That guess has a known failure mode. Polished, formulaic, or non-native writing can score as AI-written, and when a student appeals, there is nothing behind the number to re-examine.
During the work
NotAI reads how the work was produced while it happens: keystroke rhythm, pauses and corrections, paste events, and the technical signals automation gets wrong. It never judges whether the finished words sound human.
No camera, no microphone, no locked-down browser. When a session looks automated, the flag arrives with a full replay, and a human reviewer makes the call. New to the category? Read the plain-language guide to behavioral AI detection.
Around the exam
Lock down the exam environment: browser restrictions, identity checks, and camera and microphone feeds recorded for later review. A heavy-touch answer built for sealed, high-stakes exams.
The cost is friction and reach: setup before every exam, a camera in every student's room, hours of recordings to review, and nothing at all between exam windows, where most coursework actually happens.
02 · The same six questions
Swipe to compare approaches
| How they differ | AI text detectors | NotAI | Proctoring software |
|---|---|---|---|
| What it examines | The finished writing, and whether its style "sounds" AI-generated | How the work was produced: typing rhythm, pauses, corrections, paste events, and technical signals | The exam environment: camera, microphone, and screen feeds during a locked-down session |
| When it runs | After submission, once the work is done | During the work, in the background | During scheduled exams, with setup before each one |
| False positives | A style score with nothing behind it to re-examine; polished or non-native writing can score as AI | A flag is never a verdict: a reviewer replays the full session before any decision | Ordinary behavior (a glance away, background noise) generates flags for staff to sort through |
| Student experience | Invisible until an accusation arrives, backed only by a score | Nothing to install, nothing to perform; students just do the work | Room scans, identity checks, and a locked-down machine before each exam |
| Privacy surface | Finished work is sent to a third party for scoring | Typing and interaction signals on the page; no camera, no microphone, no screen recording | Video and audio from the student's room, plus screen and device access |
| Evidence produced | A percentage score | A full session replay, reviewable keystroke by keystroke | Recordings that take about as long to review as the exam itself |
03 · The honest part
Comparison pages usually end in a column of green checkmarks. This one ends with the limits, because certainty is the one thing this market has over-promised.
Not text scoring, not proctoring, and not NotAI, in either direction. Our answer to uncertainty is evidence rather than confidence: when a session merely looks unusual, the flag arrives with the full session replay, and a reviewer sees exactly what happened before any call is made.
Some AI agents will try to fake human behavior, and we do not claim evasion is impossible. Detection models update continuously against new agent behavior, and tampering with the monitoring is itself a signal NotAI looks for. The job is to keep raising the cost of faking human work, and nobody, including us, has ended the race.
Whichever category you choose, no automated system should decide a student's academic fate. NotAI is built on that assumption: flags route to human reviewers with the full session evidence, and atypical typing alone, including from students who type slowly or use assistive technology, condemns no one. See how a verdict is reached.
If a sealed, high-stakes final is your whole problem, proctoring's tradeoffs may be worth it, and the two approaches are not mutually exclusive. NotAI is built for everything else: the everyday essays, discussions, and take-home work where cameras do not belong and finished-text guessing does not hold up. What we will and will not claim about our own system is published in the AI transparency disclosure.
Walk through the pipeline from first keystroke to human review, then compare plans. Academic pricing is a flat annual fee based on institution size.