We read the writing process, not the prose.
Learnawaynever guesses whether text “looks like AI.” Instead it records how a response is entered and applies a transparent, rule-based engine: no AI model, no probability black box.
Create & share
Type your assignment name, get a link, share it. Students open the link with no account needed.
Capture the process
As students write, the browser records timing-only telemetry: event types, timing deltas, and paste lengths. Never the characters.
Score with rules
A deterministic engine turns the telemetry into a plain-English label and explanation tags, each backed by a data point you can verify.
The signals we look at
Paste dominance
What share of the final text arrived via paste, and how big each paste was.
Typing cadence
Whether keystroke rhythm is natural and varied, or robotically uniform (a sign of transcribing).
Speed vs volume
Whether the writing time is plausible for the word count.
Late insertions
Large blocks appearing near the end with no typing before them.
Tab switching
How often the student left the submission tab.
Interaction variety
Whether substantial text was produced via just one or two events.
No single signal raises concern on its own. Multiple corroborating signals are required before a submission is flagged as high concern. This protects students who legitimately paste from their own notes or use assistive technology.