Direct answer
A "written by AI" checker works by scoring text for statistical patterns typical of large language models and returning a percentage of the qualifying text it believes was AI-generated. Turnitin's AI writing detection runs inside the Similarity Report, highlights the specific segments it flags, and presents the result as an indicator for instructors rather than a verdict [1]. The percentage tells you how much of the analysed prose matched AI-like patterns — not who typed it, and not whether the student intended to cheat.
How Does an AI Writing Checker Decide Whether Text Was Written by AI?
Detection starts with a coverage question: what part of the document is even eligible for analysis. Turnitin's report only evaluates prose that meets minimum length requirements, so short answers, reference lists, headings, and tables are excluded from the calculation [2]. That means the headline percentage is always a proportion of qualifying text, not of the whole file — a 40% result on a 2,000-word essay does not mean 800 words were copied from a chatbot.
Inside that eligible text, the model looks for segments whose rhythm and word choice resemble machine output rather than human drafting. The report then highlights those segments directly in the document so a reader can inspect them one by one instead of trusting a single number [2]. This segment-level view matters because most real academic writing is mixed: a student may draft an outline with a model, rewrite it by hand, and paste in two AI-polished paragraphs.
Finally, the checker aggregates the flagged segments into the overall percentage shown at the top of the report. Because the figure is built from flagged qualifying text, it behaves as a weight, not a probability — a higher number means more of the submitted prose looked machine-like, not that the system is more certain the student cheated [2]. Reading the highlights, not just the score, is how the tool is designed to be used.
How Accurate Is a "Written by AI" Result, and What Does the Percentage Actually Mean?
No detector is a lie detector, and Turnitin publishes accuracy figures alongside an explicit warning that its AI score is a low-confidence signal at the low end of the range [3]. When confidence is insufficient — notably below the 20% threshold — the score can be displayed as an asterisk rather than an exact percentage, which is a signal of uncertainty rather than a clean bill of health [3]. Students who see an asterisk sometimes assume they have been cleared; what they actually have is a result too weak to quantify.
The percentage is best read as a share of flagged qualifying text, so it can move for reasons that have nothing to do with misconduct. Heavily templated academic phrasing, formulaic methodology sections, and text that was translated or heavily paraphrased can all produce machine-like patterns [3]. Equally, a genuinely AI-generated paragraph that a student rewrote sentence by sentence may score low. Mixed authorship produces partial percentages precisely because the underlying text is genuinely mixed.
What the number cannot tell anyone is intent. A student who used a model to tighten grammar and a student who generated an entire essay can both land in the same band, and the report offers no way to separate them [3]. That is why institutions pair the score with drafts, version history, and conversation with the student — the checker supplies evidence, and the human supplies context.
How Can a Student Verify Their Own AI Score Before Submitting?
The practical obstacle is access. Institutional Turnitin checks are controlled by the institution's assignment settings, so students generally cannot run a submission through their university's account to see the AI report in advance [4]. Many students only discover their score after the deadline has passed, when revision is no longer possible and the number is already attached to their record.
The workable approach is to review a draft against the same report format an instructor sees, then revise the specific segments that were flagged rather than rewriting the whole document. Because the report highlights passages instead of only printing a percentage [2], a student can see whether the flagged text is a formulaic methods paragraph — often fixable by adding specific detail and original phrasing — or a block of untouched model output that needs genuine rewriting.
Transparency also helps. Institutions increasingly expect students to disclose AI assistance and to keep drafts as evidence of their process [4], so a student who can show an outline, an early draft, and a revision trail is in a far stronger position than one who can only produce a final file. Verification before submission is therefore less about chasing a zero and more about knowing exactly which sentences would be questioned, and being able to explain them.
Knowing which sentences a detector would flag is only useful if you can see them before your deadline does the deciding for you. turnitin0 lets you run your draft through a pre-submission check and receive the same two PDFs your instructor's system produces — the AI detection report with flagged segments and the similarity report — usually in under 15 minutes, without adding your file to Turnitin's student paper database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does a high AI score prove a student used ChatGPT?
No. The score measures how much qualifying text matched AI-like patterns, and Turnitin itself frames the result as an indicator that should be read alongside other evidence [1]. A high percentage is a reason to look at the flagged segments, not a conclusion about authorship.
What does an asterisk instead of a percentage mean?
It means the detector's confidence was too low to display an exact figure, which typically happens below the 20% threshold [3]. Treat it as an uncertain result rather than a confirmed pass.
Can a student run their own Turnitin AI check through their university?
Usually not. Institutional checks are governed by assignment settings, so students cannot normally self-submit to preview the AI report [4]. Independent pre-submission checks exist for exactly this reason.
Do reference lists and short answers affect the AI percentage?
They are excluded. Turnitin only analyses prose that meets minimum length requirements, so bibliographies, headings, and brief responses do not inflate the score [2].
Can flagged text be revised without rewriting the whole essay?
Yes. Because the report highlights individual segments [2], most students revise the specific flagged paragraphs — adding concrete detail, original phrasing, and their own analysis — rather than rebuilding the document.