Can Turnitin Detect ChatGPT? What the Report Can and Cannot Show

Yes—if an institution has Turnitin's AI-writing feature enabled, Turnitin can flag qualifying passages that resemble text produced or altered by generative AI. But it does not inspect a student's ChatGPT account, identify a particular prompt, or prove that ChatGPT was used.

That distinction matters. The report answers a statistical question about the submitted prose. It does not answer the policy question of whether a student broke an assessment rule.

Short answer: Turnitin can identify writing patterns that its model classifies as likely AI-generated, AI-paraphrased or modified by a bypasser. Its own guidance says the model can be wrong and must not be the sole basis for action against a student.

Reading up on Turnitin's AI detection generally? This page answers the ChatGPT-specific question. The complete guide to how the AI writing report works — and whether your university runs it at all — is does Turnitin detect AI?

What Turnitin actually detects

Turnitin describes its AI-writing capability as a model for finding qualifying text that may have been produced by a large language model or altered with AI paraphrasing or bypass tools. In practical terms, the system analyses patterns in the submitted writing. It does not have access to ChatGPT logs.

As of 20 July 2026, a submission must contain at least 300 words of long-form prose and no more than 30,000 words to receive an AI-writing report. Turnitin lists English, Spanish and Japanese as supported languages, although some paraphrase and bypass detection features are English-only. Poetry, code, scripts, bullet lists and annotated bibliographies are not the model's intended material.

Turnitin also updates its model. A February 2026 release improved recall, and Turnitin notes that an old report is not recalculated automatically: resubmitting the same document after a model change can produce a report from a different model version.

What the AI percentage means

The percentage is the proportion of qualifying prose that Turnitin's model identifies as likely AI-generated or AI-altered — not the probability that the student used ChatGPT, and not necessarily a share of the whole file. Scores of 1-19% display as an asterisk because false positives are more common in that range.

The full anatomy of the report — qualifying text, the asterisk band, who can see it, and the institutional on/off switch — is covered once, in depth, in does Turnitin detect AI?. For score language across tools, read what an AI-detection percentage means.

What Turnitin cannot prove

An AI-writing report cannot establish:

• which AI product, account or prompt was used;
• who typed or edited a passage;
• whether AI use was permitted for that assignment;
• whether cited ideas are accurate;
• whether text is plagiarised; or
• whether academic misconduct occurred.

The same statistical pattern can have more than one cause. A student may have used prohibited generative AI, permitted language support, conventional editing, a template, or no AI at all. A classifier does not reconstruct that history.

Turnitin's own current guide says the model may misidentify human, AI-generated and AI-paraphrased writing. Its review guidance tells educators to use the report as one data point alongside knowledge of the student, the work and institutional policy.

AI score versus similarity score

The point that matters for ChatGPT specifically: a ChatGPT-written passage often scores low on similarity, because it copies no indexed source. The similarity report can be clean while the AI report is not — which is why "my similarity score was 3%" says nothing about AI use.

The two reports are compared side by side in does Turnitin detect AI?, and the full four-way result matrix is in AI detection versus plagiarism detection.

Copy-paste, Copilot and paraphrased text

Three variants of this question dominate related searches, so here are the direct answers:

• "Can Turnitin detect copy and paste from ChatGPT?" Pasting unedited ChatGPT output is the easiest case for the model — the statistical patterns are exactly what it was trained on. Note this is the AI report, not the similarity report: fresh AI text usually matches no source, so the plagiarism side may stay quiet while the AI side flags.
• "Can Turnitin detect ChatGPT if I paraphrase it?" Its model has a category for AI-generated-then-paraphrased text, with mixed published results — the paraphrase question has its own guide.
• "Can Turnitin detect Copilot or Gemini?" The model classifies AI-like prose without naming a tool; text from Copilot, Gemini, Claude or any other LLM is the same detection problem as ChatGPT. No report will ever say "this was Copilot".

For how the report itself behaves — the 300-word floor, the asterisk band, the institution toggle — see does Turnitin detect AI?

How educators should review a result

1. Check eligibility. Confirm language, length and prose format are suitable for the model.
2. Read the highlighted passages. Do not act on the headline percentage alone.
3. Check the assessment rule. Distinguish prohibited generation from permitted brainstorming, proofreading or declared assistance.
4. Compare process evidence. Look at outlines, notes, citations, drafts, version history and previous work where appropriate.
5. Invite an explanation. Ask neutral questions about the argument, sources and writing process.
6. Follow the institution's procedure. Share the allegation and evidence, keep records and provide a meaningful opportunity to respond.

The UK's Office of the Independent Adjudicator says interpretation of detection-software reports involves evidence-based academic judgement, and that the report and the academic analysis should be shared with the student.

If you want a second, passage-level reading of the same prose, Genutext's pay-as-you-go AI detector returns an AI-writing signal with sentence-level context in paid scans — another input for review, not a verdict.

What students should keep

The safest response is not to chase a lower detector score. Keep ordinary evidence of genuine work:

• dated notes and reading records;
• an outline that develops into the final structure;
• draft and version history;
• the sources behind factual claims;
• any required AI-use declaration; and
• the assignment guidance that applied when the work was submitted.

If AI assistance is allowed, describe it accurately. If you are unsure, ask the tutor before submission. A transparent process is more persuasive than a screenshot from a second detector.

Frequently asked questions

Can Turnitin tell that I used ChatGPT specifically?

No. Turnitin can classify passages as likely AI-generated or AI-altered, but it cannot see a ChatGPT account, conversation or prompt and cannot identify ChatGPT as the definite source.

Does a 0% score prove no AI was used?

No. A 0% result means the model did not identify qualifying text as likely AI-written in that submission. False negatives, unsupported formats and mixed writing processes remain possible.

Is a Turnitin AI score the same as plagiarism?

No. AI-writing detection classifies writing patterns; the Similarity Report finds text overlap. Neither score alone determines plagiarism or misconduct.

Can a student see the AI-writing report?

Access depends on the institution's Turnitin setup and process. Where a report informs an allegation, fair procedure should include sharing the relevant evidence and analysis with the student.

Sources and further reading

• Turnitin: Using the AI Writing Report, updated 6 March 2026
• Turnitin: File requirements for an AI writing report
• Turnitin: AI writing detection model release notes
• Turnitin: How should I review the AI Writing Report?
• OIA: Academic disciplinary procedures

Turnitin is a trademark of Turnitin, LLC. Genutext is not affiliated with or endorsed by Turnitin.