Paste a student's work.
See exactly where it breaks down.
Made for busy professors: paste the text, click once, and get line-by-line flags for AI generation, plagiarism, citation errors, auto-correct artifacts and translation tells. No download, no account, nothing to set up.
What the scan looks for
Every check is a transparent rule you can read and judge yourself — no black box. Here is how each category works, with real examples and what to do with a hit.
AI generationhow it works + examples + tips▾
How it's detected
Compares the text against 100+ phrases that language models over-use, leftover chatbot boilerplate, and the shape of the writing itself — sentence rhythm, vocabulary spread, paragraph uniformity and punctuation habits. It also spots “humanized” AI text: polished machine prose with fake errors sprinkled in to look human.
Example flags
- It is important to note that… — Hedging opener that appears in model text far more often than in student writing.
- delve into — A signature model word — its use exploded after chatbots became widespread.
- Certainly! Here's a revised version: — Assistant reply text pasted straight into the document — near-certain proof of chatbot use.
- In today's fast-paced world… — The single most common generated-essay opener.
Tips — what to look for
- Look for text that is polished but empty — flawless grammar with no specific detail, anecdote or course reference.
- Every sentence the same length is a tell: human writing mixes short punches with long rambles.
- Watch for “Firstly, secondly, finally” marching order and a conclusion that merely restates the intro.
- Check the document's invisible characters — paste it in; invisible watermark characters show up here.
- The strongest evidence is boring: leftover chat artifacts like “I hope this helps” or unfilled [insert topic] placeholders.
- Beware the “humanizer” trick: flawless, formulaic prose with a few random typos, a lowercase “i”, or slang like “tbh” dropped in. Real student errors are spread evenly — fake ones sit on top of otherwise perfect text.
Plagiarism suspicionhow it works + examples + tips▾
How it's detected
Spots phrasing typical of copied sources — encyclopedia definitions, homework-site references, unattributed claims and publishing residue. It cannot search the web, so treat hits as leads to check, not proof.
Example flags
- Studies show that… — An appeal to research with no source named — ask the student which study.
- is defined as — Textbook-definition phrasing, often lifted verbatim from a reference work.
- according to Wikipedia — Names a summary site instead of a scholarly source.
- all rights reserved — Copyright boilerplate left behind by a copy-paste from a published page.
Tips — what to look for
- Paste any suspicious sentence into a search engine inside quotation marks — copied passages surface immediately.
- A sudden jump in vocabulary or register mid-essay is the oldest tell; compare it to the student's earlier work.
- Ask the student to explain one flagged sentence in their own words, live — authors can, copiers usually can't.
- Check the reference list against the text: copied work often cites sources the essay never actually uses.
Citation errorshow it works + examples + tips▾
How it's detected
Checks the mechanics of sourcing: whether in-text citations have matching entries, years are plausible, page numbers are attached where required, and quotes are properly framed.
Example flags
- (Smith) — Author with no year — incomplete citation in most styles.
- et al. argued… — “Et al.” with no year and possibly no matching reference entry.
- (p. 42) — A bare page number with no author attached.
Tips — what to look for
- Spot-check one or two citations against the actual source — fabricated references often look perfect on paper.
- AI tools invent realistic-looking citations; if a source doesn't exist in your library search, that's the finding.
- Citations clustered only at paragraph ends, never mid-argument, often means they were sprinkled on afterwards.
Auto-correct artifactshow it works + examples + tips▾
How it's detected
Finds the wrong-word substitutions, merged words and agreement slips that phone and editor autocorrect leave behind — signs of rushed typing or last-minute editing rather than AI use.
Example flags
- defiantly — The classic autocorrect of “definitely”.
- pubic health — A one-letter autocorrect of “public” — always worth a tactful note.
- could of — Speech-to-text or autocorrect turning “could've” into “could of”.
Tips — what to look for
- These flags are about care, not misconduct — many suggest hasty typing rather than cheating.
- Heavy autocorrect artifacts plus perfect AI-style phrasing elsewhere can mean parts were typed and parts generated.
Machine translationhow it works + examples + tips▾
How it's detected
Detects literal-translation fingerprints: wrong prepositions, pluralized uncountable nouns, doubled modals and false friends borrowed from the student's first language.
Example flags
- according to me — A literal translation; idiomatic English says “in my opinion”.
- informations — Uncountable noun pluralized — one of the most reliable translation tells.
- more better — A double comparative produced by translating word-for-word.
Tips — what to look for
- Translation tells often appear alongside flawless passages — a sign parts were written, then machine-translated.
- Compare with the student's spoken English in class; a wide gap is worth a conversation, not an accusation.
- Ask which tools are allowed in your policy first — many institutions treat translation help differently from AI drafting.
Feeling the love?
… educators helped
Be the first — one click adds you to the educators helped tally.
your personal tally — saved in this browser only