Direct answer
An "X" (or a crossed/dashed indicator) on an AI-generated content detection result means the detector could not assign a confident AI percentage to that section, not that your work has been proven to be machine-written. Turnitin reports AI writing as an overall percentage plus highlighted segments, and it replaces the number with an asterisk (*%) when its confidence falls below its own threshold [1]. Treat the mark as a low-confidence signal that needs a human read, not as a verdict.
What does the "X" mean on a Turnitin AI-generated content detection result?
The single most useful thing to understand is that Turnitin's AI writing report is built from two layers: an overall percentage and a per-segment highlight layer. The overall figure summarises how much of the submitted text the model classifies as AI-generated, while the segment layer marks the specific passages that drove that figure [2]. When a passage is shown with a placeholder mark instead of a number, you are looking at the confidence layer, not the accusation layer.
Turnitin explicitly frames "detected" text as text its model classifies as AI-generated, and it does not claim that classification is proof of misconduct [2]. The detector reads statistical patterns, so its output is probabilistic by design. A mark that reads as an X, a dash, or an asterisk is the report telling you the evidence in that segment was too weak or too mixed to quantify.
The asterisk convention is the one students most often misread. Turnitin displays *% instead of an exact figure when AI detection falls below its 20% confidence threshold, which means the system is reporting a low-confidence signal rather than a measured score [1]. So an X-style mark and a *% are close cousins: both say "something was noticed, but not enough to stand behind a number."
In practice, the report your instructor opens inside the LMS is the same report you would see in a preview, including the segment highlighting and the threshold behaviour [1]. That matters because it means you can rehearse the exact reading experience before it counts against you, and you can see which segments produced the ambiguous mark.
Why does Turnitin flag AI-generated content as detected when I wrote parts myself?
Because the detector measures how text is built, not who built it. Turnitin's model looks at patterns of word choice, sentence rhythm, and structural predictability across a passage; it has no record of your drafting history, your notes, or the hours you spent on a paragraph [3]. Any stretch of prose that happens to look statistically predictable can be pulled into the flagged set, even when a human typed every word.
Formulaic academic writing is the classic trigger. Templated transitions, balanced tricolons, hedged topic sentences, and the polished cadence produced by grammar and paraphrasing tools all resemble the patterns the model associates with machine generation [3]. Literature reviews and definitional openings are especially exposed because the genre itself rewards conventional phrasing.
Turnitin also acknowledges that heavily edited drafts and writing from non-native English speakers are more likely to be mischaracterised, since both can flatten the natural variance the detector uses as a human signal [3]. If you wrote the argument yourself and then ran it through a rewriting tool, you may have smoothed away exactly the irregularities that would have read as human.
The practical consequence is that the score cannot separate "AI-assisted" from "AI-generated" [3]. A detected flag is a prompt to look at the highlighted segments and judge whether the phrasing is genuinely yours, not a statement about your intent.
How can I lower a detected AI score before I submit?
Start with the flagged segments rather than the whole document. Rewrite those passages at the level of argument structure, not vocabulary: rebuild the claim, insert your own reasoning step, and let the sentence order follow your thinking instead of a template [4]. Synonym swapping and light paraphrasing leave the underlying predictability intact, which is why scores often barely move after a superficial edit.
Then add material a model could not have produced. Specific evidence, precise citations, named sources, and first-person analysis of a dataset or case all introduce the kind of concrete irregularity that reads as authored [4]. Concrete detail is also what makes the prose better, so this step improves the work and the report at the same time.
Vary the surface of your writing as well. Mix short declarative sentences with longer qualified ones, remove the connective scaffolding that appears in almost every AI draft, and cut any sentence that exists only to announce what the next sentence will do [4]. Read the result aloud; anything you would not say to a tutor in a supervision meeting is a candidate for rewriting.
Finally, re-check before you submit, and check in a way that does not add your file to Turnitin's student paper database. A non-repository preview returns the same AI and similarity reports your instructor sees, so you can confirm whether the flagged segments actually cleared [4]. Turnitin0 delivers both PDFs together in under 15 minutes in 98% of cases, which fits inside a normal revision window.
If the rewrite still leaves flagged segments and your deadline is close, turnitin0's AI humanizer is the faster route: it rewrites the flagged passages while preserving your meaning, citations, headings, and.docx formatting, then re-checks the result against Turnitin.
※ Turnitin0.com - AI Humanizer Bypassing Turnitin [AI Detector](https://www.turnitin0.com/)
FAQ
Is an "X" on an AI detection report the same as being accused of cheating?
No. The mark reflects the detector's confidence in a segment, not a finding of misconduct; Turnitin describes detected text as text its model classifies as AI-generated and does not present it as proof [2]. Your institution's policy, not the symbol, determines what happens next.
Does *% mean my AI score is zero?
No. Turnitin shows *% in place of an exact figure when AI detection falls below its 20% confidence threshold, so it signals a low-confidence result rather than a measured zero [1]. Ask your tutor how they read threshold results before you assume the worst.
Can a fully human-written paragraph be flagged?
Yes. Because detection is pattern-based, formulaic academic phrasing, grammar-tool polishing, and conventional literature-review prose can all be pulled into the flagged set even when you wrote every word [3].
Will rewriting flagged passages always clear the score?
Not always on the first pass, which is why a re-check matters. Structural rewriting plus added specific evidence and citations moves scores most reliably, while synonym-level edits often leave them unchanged [4].
Can I check my draft without it entering Turnitin's student paper database?
Yes, with a non-repository preview service. Turnitin0 checks.docx,.pdf, or.txt files between 300 and 30,000 words, returns the AI and similarity PDFs together, and does not add your file to the student paper database or share reports with third-party databases.