Direct answer
A "controllo AI in un testo" is the process of running a finished draft through an AI writing detector to see what share of it reads as machine-generated. Turnitin does this inside its similarity report: it returns an overall AI percentage plus a sentence-by-sentence highlight layer, and only text of 300 words or more is eligible for scoring [1]. The number is a probabilistic estimate, not a verdict, so the useful move is to read it alongside the flagged sentences rather than on its own [1].
How Do You Check A Text For AI Writing, And What Does The AI Score Mean?
The check itself is straightforward: the document is submitted, the detector segments the prose into sentences, and each sentence is classified as AI-generated, AI-paraphrased, or human-written. The overall figure you see is the proportion of qualifying text that fell into the AI categories, which is why a 34% score never means "34% of the essay is fake" but rather "34% of the scored sentences carried AI signals" [2]. Because the report is generated from the same engine instructors use, a preview opened before submission shows the same layout and the same sentence-level flags your marker will see in the LMS inbox [2].
What surprises most students is how much of a document is not scored. Titles, reference lists, quotations, and short submissions are excluded from the calculation, so a heavily cited literature review can score low even when the body prose is entirely machine-drafted — and a short reflective piece under the word minimum returns no AI figure at all [1]. Read the denominator before you read the percentage.
Treat the score as a diagnostic pointer rather than a grade. Turnitin itself frames the indicator as an estimate of likelihood that should never be the sole basis for an academic integrity decision, which means the defensible question is not "what number did I get" but "which sentences produced it, and can I account for how they were written" [1]. That distinction is what turns a nervous score-check into actual preparation.
Why Does Turnitin Show An Asterisk Instead Of A Percentage?
When the AI score sits below Turnitin's 20% confidence threshold, the report prints *% in place of an exact figure. The asterisk is a confidence label, not a clean bill of health: it tells you a small share of the text scored as AI and the signal was too weak to state precisely, so the detector declines to commit to a number [3]. A document showing *% is not certified as human-written — it is simply one where the evidence was thin.
This matters because students routinely misread the symbol in both directions. Some assume *% means the detector failed and ignore it; others panic and start rewriting prose that was never flagged. The honest reading is narrower: the overall figure is low-confidence, while the sentence-level layer still shows exactly which passages attracted attention, and that layer remains available whether or not a precise percentage is printed [3].
Scores at or above the threshold are displayed as whole percentages, and those are the cases where the number deserves real scrutiny. If you see a double-digit figure, open the sentence view first — the flagged spans usually cluster in a few predictable places, such as transition-heavy topic sentences, generic literature summaries, and paragraphs that restate a source without quoting it [3]. Locating the cluster tells you far more than the headline number does.
How Can You Lower A High Turnitin AI Score Without Losing Meaning?
The effective method is to revise the sentences that were actually flagged, in your own voice, with your own evidence. Turnitin's own guidance to institutions emphasises that students should rework flagged passages themselves — restating the argument in their own phrasing, adding specific detail and analysis, and leaving citations intact — rather than running a thesaurus over the text, which changes surface wording while preserving the statistical fingerprint that triggered the flag [4]. Cosmetic synonym swaps are the single most common reason a second submission scores the same as the first.
Work sentence by sentence rather than document-wide. Because the score is a proportion of scored sentences, rewriting the specific flagged spans moves the number far more reliably than editing paragraphs that were never highlighted [4]. In practice that means opening the sentence view, listing the flagged sentences, and for each one asking what you actually wanted to claim — then writing that claim with a concrete example, a figure, or a citation attached. Specificity is what human academic prose has and generic AI prose lacks.
Finally, keep a record of the drafting process. Since the indicator is explicitly not designed to stand alone as proof of misconduct, the strongest position a student can hold is being able to explain how a passage came to exist — notes, earlier drafts, source annotations [4]. A revision that lowers the score and leaves you able to defend the text is worth more than a number that merely looks better.
Reading the flags is the easy half; the harder half is knowing, before the deadline, whether your revised draft actually clears them. That is exactly the gap turnitin0 closes: you upload your .docx, .pdf, or .txt and receive the same two PDFs your institution generates — the AI detection report and the similarity report — in under 15 minutes in 98% of cases, checked without adding your file to Turnitin's student paper database. With 100,000+ reports delivered to 20,000+ students and a 4.9/5.0 satisfaction rating, it is the practical way to run a controllo AI on your own text and see the sentence-level flags before anyone else does.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is a controllo AI the same thing as a plagiarism check?
No. A similarity check compares your text against a source database and reports matches; an AI check estimates whether the prose itself reads as machine-generated. Turnitin produces both from one submission, and they are separate reports with separate scores [1].
Does Turnitin check every word of my document?
Only qualifying prose of 300 words or more is scored. Quotations, references, and very short submissions are excluded, which is why the percentage always refers to a subset of your text rather than the whole file [1].
What does *% mean — is my document clean?
It means the AI score fell below Turnitin's 20% confidence threshold, so no precise figure is printed [3]. That is a low-confidence signal, not a certification that the text is human-written.
Can I lower the score by paraphrasing with a thesaurus?
Rarely. Surface synonym swaps preserve the underlying statistical pattern that triggered the flag; rewriting the flagged sentences with your own analysis and evidence is what actually moves the number [4].
Will my instructor see that I checked my draft first?
Not from the report itself. A pre-submission check run through an independent service is not added to Turnitin's student paper database and is not shared with your institution, so the report your instructor generates comes from your own final submission.