AI detection that respects how high the stakes are
AI writing detection is more than spotting one punctuation mark or a familiar phrase. Checkmark analyzes the relationships among grammar, word choice, sentence structure, and rhythm, then shows which passages strongly resemble generated writing and how the text was actually typed.
Passage-level, not paper-level
AI-flagged passages are underlined in the essay itself, each with its own evidence card - you see which parts, not just how much.
Calibrated confidence
Each finding shows how strongly a passage resembles patterns common in AI-generated writing - a high-confidence signal, not a claim about authorship.
Backed by process evidence
The same report shows paste events and typing patterns, so an AI signal can be confirmed - or cleared - by how the text was written.
A signal, not a verdict
Every AI finding renders as a gradient slider between typical human-writing patterns and typical AI-writing patterns. It communicates the strength of the signal without pretending that a finished passage can prove who wrote it:
Typical AI writing pattern versus typical human writing styles. Do not solely rely on this score to determine AI authorship.
This disclaimer appears on every AI Detection card.
We put that in the product because it's true - of every AI detector, including ours. What makes the signal useful is the evidence around it.
Guardrails we build in
Short texts say N/A
Below ~150 words there isn't enough signal for a trustworthy AI-writing finding, so the report says N/A instead of guessing.
Findings start conversations
Flag statuses (Flagged, Resolved, Not Flagged) are teacher-only. Nothing is shown to a student as an accusation.
Cross-checked by process
A passage that strongly resembles generated writing but was visibly drafted, revised, and typo-corrected over 40 minutes tells a different story than one pasted in at once. The report shows both signals.
Why detection alone was never enough
AI detectors - all of them - work on the finished text. Humanizer tools exist specifically to launder that text. But no tool can launder the writing session itself: a 900-word essay that appeared in one paste event, or was transcribed line-by-line at machine-steady speed, looks nothing like authentic drafting. That's why every Checkmark AI-writing finding ships inside a report that also reconstructs the writing process.
Judge it on your own essays
Paste real student writing (or ChatGPT output) into the demo and see the full report in about a minute.