If your professor showed you a Turnitin report saying some percentage of your paper is AI-generated, you are looking at an estimate, not evidence of what you did. The score comes from a statistical model predicting whether text resembles AI writing, it cannot identify where any sentence actually came from, and the company behind it acknowledges it gets papers wrong. Schools treat that number very differently, some as a conversation starter and some as proof, and which treatment you get often depends less on the report than on how you respond to it.
What Is the Turnitin AI Score, and What Is It Not?
Turnitin’s AI detection assigns scores to segments of text based on how closely they resemble patterns of AI-generated writing, then reports an overall percentage. That is a probability judgment about style, not a record of authorship. There is no database of AI outputs being matched, no log of your writing session being read, and no way for the tool to know whether you used ChatGPT, wrote at 2 a.m. in a formal mood, or simply write the way the model associates with machines.
This is not a technicality. It is why the tool’s own maker has publicly acknowledged real-world error, estimating at one point that about 1 percent of documents and 4 percent of sentences were incorrectly flagged, and why universities including Vanderbilt, Michigan State, Northwestern, and the University of Texas at Austin turned the feature off rather than risk accusing innocent students. Some of the most damaging demonstrations came from accused students themselves, who showed detectors flagging historic human writing as machine-made.
A percentage is where a school’s inquiry is supposed to begin. The problem is that at many schools, it is where the inquiry effectively ends, unless the student changes that.
Is There a Percentage That Counts as Cheating?
No. There is no threshold at which a Turnitin score becomes a finding, no standard that says 40 percent is innocent and 80 percent is guilty, and no policy language anywhere that converts the number into responsibility on its own. What matters is what your school’s process does with it: whether the professor treats it as one factor or the whole case, what other observations get attached to it, and what standard of proof the hearing applies. Most schools decide these cases on whether misconduct is more likely than not, which means an unanswered flag paired with a professor’s impression can be enough, and a well-answered flag frequently is not.
That asymmetry is the entire game. The score does not decide the case. The response does.
What Should You Find Out Before Responding?
Three things shape everything, and students rarely check any of them. First, what exactly you are accused of, because “AI was used” and “unauthorized assistance” and “work not your own” are different charges with different defenses. Second, what your school’s policy actually says about AI and about detection tools, because some schools have formally cautioned against relying on detector scores, and a case built on one at such a school has a structural problem worth knowing about. Third, what the full report shows, not the summary number, because which sentences were flagged, and whether they are the ones containing your actual analysis or just formulaic transitions, changes what the report is worth.
Knowing what to look for in those three places, and what to do with what you find, is where these cases are genuinely won, and it is the part no article can responsibly script, because it depends on your school, your policy, and your paper.
What Makes Turnitin Cases Go Wrong for Students?
Arguing with the number instead of the case. Students fixate on proving the percentage wrong. The stronger question is what the school can actually establish about your paper, which is a different fight with better ground.
Assuming the professor understands the report. Many instructors are reading these reports for the first time too, and misreadings get built into accusations. Left uncorrected, a misreading becomes the record.
Producing a new writing sample on demand. Some students are asked to rewrite the paper or write on the spot to prove themselves. Agreeing without understanding how it will be judged, and by whom, has sunk cases that the original evidence never could have.
Treating the meeting as informal. The conversation where you are shown the report is the fact-gathering stage of a formal process, whatever it is called.
How Does an Attorney Handle a Turnitin Flag?
Richard Asselta is a student defense attorney who defends students in AI detection cases and academic misconduct matters nationwide. He analyzes what the school’s evidence actually establishes beyond the score, measures the case against the school’s own written policies on AI and detection, prepares the response and hearing presentation, and makes sure the report is treated as what it is rather than what the accusation needs it to be.
Because these cases run under each school’s own code of conduct, he defends students at colleges and universities across the country.
Frequently Asked Questions
My Turnitin AI score is very high. Does that mean I will be found responsible?
A high score raises the stakes but decides nothing by itself. Cases with high scores are won on authorship evidence and policy grounds, and cases with modest scores are lost on bad responses. The score sets the starting point, not the outcome.
Am I entitled to see the full Turnitin report?
You should ask for it, and many schools will provide it as part of the process. What the sentence-level detail shows is often more useful to the defense than the summary number, which is exactly why reviewing it properly matters.
My school says it does not use AI detection, but my professor used a different detector anyway. Does that help me?
It can matter significantly. A case built on a tool the institution has declined to endorse raises questions about the school following its own standards, and that is worth developing correctly rather than mentioning in passing.
Can I be flagged for work I wrote with permitted AI help?
Yes. Detectors cannot distinguish permitted use from prohibited use, so students who followed the syllabus get flagged alongside those who did not. What was permitted, and how your use fit it, becomes the center of the case.
The Score Is an Estimate. The Finding Does Not Have to Follow It.
A Turnitin flag starts a process, and processes are shaped by the people who engage with them. If your paper has been flagged as AI-generated, what happens next depends on choices being made right now. Attorney Richard Asselta defends students in AI detection and academic misconduct cases at colleges and universities nationwide. Call 855-338-5299 before you respond to the report.