AI Checke

An AI checker measures statistical patterns in your text — perplexity and burstiness, token predictability, and the uniformity of sentence construction — then converts those patterns into a probability-like score. It does not look up a database of AI-written sentences. It asks, in effect, how predictable your prose is to a language model, and how evenly that predictability is distributed across the document [2]. That is why a checker can flag a paragraph you wrote yourself in a flat, formulaic register, and why it can miss AI text that has been heavily rewritten.

Different tools give different numbers because they were trained on different corpora, use different segment sizes, and apply different thresholds before they will call something AI. Turnitin's own published analysis of false positives is candid that the error rate is not uniform across student populations; it is concentrated in text written by non-native English speakers and in passages that have been edited repeatedly, where the natural variation of human prose gets smoothed out [2]. A tool that reports 40% and another that reports 8% on the same file are not necessarily contradicting each other — they are answering slightly different questions with different calibration.

The practical takeaway is to stop shopping for the checker that gives you the number you want and start asking which checker your institution uses. If your university runs Turnitin, then a Turnitin score is the only one that matters for your submission, because it is the one attached to your file in the learning management system [2]. Third-party detectors are useful for a rough second opinion, but they are not the score your marker opens.

Turnitin displays an asterisk in place of a number when the amount of qualifying text flagged as AI-generated falls below the 20% threshold of the document. In that band the system treats the detection as a low-confidence signal rather than a reliable percentage, so it refuses to print a figure that would imply more precision than it has [3]. The asterisk is not a zero, and it is not a clean bill of health — it means "some AI-like text was detected, but not enough for us to stand behind a number."

This distinction matters enormously when you are reading your own report. A student who sees % and concludes "I'm fine" is reading the symbol backwards; a student who sees % and concludes "I've been caught" is doing the same thing in the other direction. The correct reading is that the report found a small flagged region, and the flagged segments themselves — highlighted in the document view — are where you should look [3]. Turnitin highlights the specific passages that drove the score, so you can judge whether the flag is a false positive on a formulaic methods paragraph or a genuine stretch of unedited AI output.

The 20% threshold also explains a common frustration: a draft that has been only lightly touched by AI often lands in the asterisk band, while a fully AI-drafted essay lands well above it. If you are sitting at *% and your institution treats any detection as a concern, the sensible move is to revise the flagged segments rather than argue with the symbol [3]. Read the highlighted text, rewrite it in your own voice, and re-check.

Draft-stage checking is the single most effective habit for avoiding a nasty surprise at submission. When you check a draft before it is graded, you see the same two things your instructor sees — a similarity report showing matched sources and an AI writing report showing flagged segments — while you still have time to act on them [4]. The reports are far more useful read segment by segment than read as a headline number, because the headline tells you there is a problem and the segments tell you where it is.

One detail worth understanding is the difference between repository and non-repository checking. A standard institutional submission can add your paper to a student paper database, which is exactly what you want to avoid for a draft you intend to revise and resubmit. Non-repository checking runs the comparison without retaining your file in that database, so your draft does not become a future match against itself or against another student [4]. If you are checking multiple drafts of the same assignment, this matters.

The workflow that works is unglamorous: finish a draft, check it, read the flagged segments, revise them, check again. Students who do this once or twice per assignment stop being surprised by their final reports, because they have already seen the report and fixed what it pointed at [4]. It also builds a more honest sense of your own writing — you start to notice the flat, over-regular sentences that detectors react to, and you stop producing them.

Reading a report and fixing what it flags are two different skills, and turnitin0 exists to make the first one fast. Turnitin0 is an independent service, not affiliated with Turnitin, LLC, and it gives you the same two PDFs your professor opens in the LMS — an AI detection report and a similarity report — on a draft you upload, with English documents between 300 and 30,000 words and a turnaround that averages under 15 minutes in the large majority of orders. The check is non-repository, so your draft is not added to the student paper database, and you can delete your files afterwards. Over 100,000 reports have been delivered to more than 20,000 students across the US, UK, Canada, Australia, New Zealand, and Ireland, at a pay-per-use rate of $3.80 for a single check with no subscription.

※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary

Get Real Turnitin AI & Similarity Report