Does Turnitin Check for AI or Just Plagiarism? Both, Separately

What Does Turnitin's Similarity Report Actually Measure?

The similarity report is Turnitin's original product. It compares submitted text against a database that includes billions of web pages, academic journal articles, student paper archives, and books. When a passage matches content already in that database, Turnitin highlights the overlap and lists the source. The overall similarity percentage is simply the proportion of the submission that matched something in the database. A high similarity score does not automatically mean a student plagiarized. Properly quoted and cited text will still appear in the report. Common phrases, technical terms, standard definitions, and assignment instructions that were copied into the submission all add to the count. Instructors are expected to open the source list and evaluate whether the flagged matches are genuine plagiarism or legitimate overlap. A well-cited paper can show a 30% similarity score with nothing wrong, while a paper with a 5% score could still contain a fabricated source or an uncited paraphrase.

When Did Turnitin Add AI Detection?

Turnitin launched its AI writing detection feature in April 2023, responding to widespread use of ChatGPT in academic writing after OpenAI released the model publicly in late 2022. The AI writing indicator uses a different methodology from the similarity report. Rather than matching text against a database of known sources, it analyzes statistical patterns in the writing itself. Models like GPT-4 produce text with relatively predictable word sequences because they assign probabilities to each next token. Human writing tends to be less predictable and more variable in sentence structure and vocabulary. Turnitin's model was trained to distinguish these patterns, and the output is a percentage representing how much of the submission appears to have been AI-generated. Institutions must enable the AI indicator separately. Not every school or instructor has turned it on, which is why some students see only a similarity report.

Turnitin's AI indicator launched in April 2023 as a separate product from the similarity report, and institutions must opt in to activate it.

Are the Similarity Report and AI Indicator the Same Thing?

No, they measure entirely different things through different methods. The similarity report looks outward: it compares submitted text against existing sources in a database. The AI indicator looks inward: it analyzes the statistical structure of the writing to estimate whether a language model produced it. A student could have a low similarity score and a high AI score, meaning the writing appears original but reads as though a model generated it. The reverse is also common: a paper that closely follows a textbook or an assigned reading may score high on similarity but low on the AI indicator because the student read and paraphrased heavily in their own voice. The two reports require separate interpretations. When someone asks does turnitin check for ai or just plagiarism, the correct framing is that Turnitin checks for both, but the evidence from each system applies to different questions.

Can a Document Score High on Both Reports?

Yes, but the combination means different things than either score alone. A document with a high similarity score and a high AI score might suggest that a student used AI to generate passages and then inserted them without citation, creating matches with sources the model had seen during training. A document with high similarity and low AI would point toward conventional copying or excessive quoting. A document with low similarity and high AI is the scenario most students worry about: original-sounding but AI-generated prose. A document with both scores low is what a typical well-written, well-cited paper looks like. None of these combinations is a definitive verdict. A high AI score, for example, does not rule out the possibility that the student wrote the passage themselves and the detector made a false positive. Turnitin's own guidance states that the AI indicator is not designed to be used as the sole basis for an academic integrity decision.

Why Is the AI Score Not Proof of Misconduct?

The AI score is a probabilistic estimate, not a forensic fact. Turnitin targets a false positive rate of 1% at a 20% detection threshold, meaning that for every 100 genuinely human-written papers, roughly 1 might still be flagged. At the sentence level, the margin of error increases. Several writing patterns produce elevated AI scores for fully human work. Formal academic style with passive voice, impersonal tone, and disciplined transitions can resemble model-generated prose. Non-native English writers who follow grammar rules carefully and avoid idioms may see higher scores. Short, declarative sentences in a technical summary can also trigger the indicator. Template-heavy writing in business, legal, and scientific fields often looks like AI output to a statistical model. This does not mean the indicator is useless; it means the score should be treated as one data point that starts a conversation, not as evidence that ends one. Instructors who use it responsibly ask the student about the assignment process, review earlier drafts, and compare the work against other writing samples.

The AI writing indicator is designed to help start a conversation, not to serve as the final word on academic misconduct.

How Should Students Respond to a High AI Score?

A high AI score is stressful, but it is not the same as an accusation. The most useful response is to document your writing process before you need to. Keep your browser research tabs, outline drafts, revision history in Google Docs or Word, notes from lectures, and any rough drafts you wrote before the final version. These materials can show that the paper developed over time through your own thinking. If you receive a concern from your instructor, respond calmly and offer to explain your process. Describe the sources you consulted, the argument you developed, and the choices you made about structure and wording. If the concern goes further in a formal process, ask whether there is an appeal path and whether the standard of evidence requires more than the AI score alone. Most academic integrity policies include procedures that require meaningful evidence and an opportunity to respond.

1. Save research notes, outlines, and all draft versions before submitting.
2. Use version history in Google Docs or Word to show how your paper developed.
3. If flagged, request a conversation rather than waiting for a formal charge.
4. Ask whether the school policy requires the AI score to be combined with other evidence.
5. Keep a record of your sources and any feedback you received during the writing process.

Does Turnitin Check for AI or Just Plagiarism When Both Features Are Active?

When an institution has enabled both the similarity report and the AI writing indicator, Turnitin runs both checks on the same submission and presents them as separate sections of the report. Instructors see a similarity percentage with source citations and an AI percentage with sentence-level highlighting. The sentence highlights in the AI report show which parts of the paper the model considers most likely to be AI-generated, while the similarity highlights show which parts match specific sources. These two highlight layers are independent. A sentence highlighted in both reports would need two separate explanations. The AI score in this combined view does not affect the similarity score, and neither report adjusts based on the other. So when people ask does turnitin check for ai or just plagiarism, the practical answer in a fully enabled setup is that both checks run simultaneously but produce separate, non-overlapping evidence that an instructor must evaluate independently.

What Can Students Do Before Submitting to Reduce Risk?

Running a pre-submission check through an outside tool can give students a rough sense of how their writing reads to an AI detection model before the formal report exists. These tools are not the same as Turnitin and cannot predict its exact output, but they can flag passages that appear statistically smooth or generic. If an outside tool highlights a paragraph, read it and ask whether it makes a specific claim backed by a named source, whether the sentence structure varies, and whether it sounds like your reasoning or like a summary anyone could write. Revising toward specificity and voice is more productive than rewording to trick a model. For the similarity check, standard citation practice handles most risk. Quoting directly when you use an exact phrase, paraphrasing in your own words for everything else, and citing every source you consulted keeps the similarity score interpretable. The goal is not a low score; the goal is that every match in the similarity report can be explained as legitimate.

1. Run an outside AI detection check to find overly smooth or generic passages.
2. Revise flagged paragraphs toward specific claims, named sources, and your own reasoning.
3. Review your citations to ensure every paraphrased idea has a source attached.
4. Compare your final draft against your outline to check that your argument is still present.
5. If you used AI tools at any point, follow your school's disclosure policy before submitting.

Where NotGPT Fits in This Picture

NotGPT is a mobile AI detection tool that checks text for AI-likeness and images for AI generation. It is not a plagiarism checker and does not replicate Turnitin's similarity database. Its role in this context is as a pre-submission reading aid. Students can paste a draft into the app to see which passages score high on the AI probability scale, then decide whether those passages need revision. The Humanize feature offers a way to rewrite passages that read too generically, though the better use is to add concrete specifics rather than to mask the style. For educators, NotGPT can provide a quick second opinion when a paper is under review, helping to distinguish whether a flagged pattern is consistent across multiple tools. Neither report from any outside tool should replace the conversation and evidence review that formal academic integrity processes require. The question does turnitin check for ai or just plagiarism has a practical answer: both, separately, with independent evidence that requires human judgment to interpret.