Can Turnitin detect AI — and how do I use AI without breaking academic integrity rules?

Yes, but not the way most students fear. If you're asking can Turnitin detect AI, Turnitin does have an AI writing indicator that shows an overall percentage of the document that AI writing tools may have generated or modified, and only instructors and administrators can see it (Turnitin). Turnitin also says it does not make a determination of misconduct: the percentage should not be the sole basis for action or a definitive grading measure, and the final decision rests with the reviewer or instructor applying their institution's academic policies (Turnitin). So the tool answers a question about your text; it does not decide your case.

Can Turnitin detect AI well enough to stand on its own? Turnitin's own limits say no, and those limits matter more than the number.

The percentage is narrow. It covers qualifying long-form prose sentences only — not lists, bullet points or other non-sentence text — so it is not necessarily the percentage of the entire submission (Turnitin).
Low scores are deliberately hidden. No score or highlights are given for results between 1% and 19%; those show as an asterisk instead (Turnitin).
False positives are expected. Turnitin aims to keep its document-level false positive rate under 1% for documents with over 20% AI writing — roughly one fully human-written document in every 100 could be flagged. Turnitin lists text with little structural variation, repetitive text, and text paraphrased without developing new ideas as examples that can trigger this (Turnitin).
You can't see it. The indicator and report are not visible to students, though instructors can download the AI report as a PDF and share it (Turnitin).
Grammar checkers behave differently from generators. Turnitin says grammar, spelling and punctuation fixes from tools like Grammarly were in most cases not flagged in its tests, but content produced with generative features such as drafting, paraphrasing or summarising will likely be flagged (Turnitin).
Running text through a paraphraser or "humaniser" is not a way out. Turnitin states its detection also looks for AI-generated text modified by AI paraphraser or bypasser tools in English submissions, and that check runs automatically where AI detection is enabled (Turnitin).

Universities treat a score as a starting point, not a verdict. The University of Melbourne tells students that an AI detection result is a prompt for further investigation, that students may be asked to explain their work or provide drafts, and that a detection report alone is not sufficient evidence for an allegation (University of Melbourne). The University of Sydney says the same in different words: Turnitin's AI detection tool may be used, but the detector score would not be the only evidence relied upon for an academic integrity case (University of Sydney).

Tools that work

The tools worth using here are general AI assistants for planning, explaining and drafting — the parts your unit outline may allow. Prices and limits below come from each provider's own pages and can change; check the current plan on the tool's site before you pay.

Tool	Free tier	Paid plans shown by the provider
ChatGPT	Free plan with unlimited text chats, and limited messages with uploads, image creation, voice, deep research and memory	Plus at USD 20 per month, billed monthly, with no annual billing option; Go and Pro also listed (OpenAI)
Claude	Free at $0: chat on web, desktop and mobile, web search, file creation and code execution, memory and Artifacts	Pro at $17 per month with the annual discount ($200 billed up front) or $20 if billed monthly; Max from $100 per month (Anthropic)
Gemini	Gemini app free at $0/month with a Google Account	Google AI Plus at $4.99/month; Google AI Pro at $19.99/month with 4x higher usage than Free, Gemini in Gmail and Docs, and 5 TB storage. Prices differ by country (Google)

For brainstorming, outlining and having a concept explained, a free tier is usually enough — you are not paying for more words, you are paying for higher limits when you hit them. File handling differs: ChatGPT accepts up to 512MB per file with Free users limited to 3 file uploads per day and users generally up to 80 files every 3 hours (OpenAI), while Claude accepts up to 500MB per file and up to 20 files per chat, with PDFs up to 1000 pages (Anthropic).

Two habits matter regardless of which tool you pick. If you don't want new ChatGPT conversations used to train OpenAI models, turn off "Improve the model for everyone" under Settings > Data controls — your chats still stay in history (OpenAI). And treat every output as unverified: OpenAI says ChatGPT can produce incorrect or misleading outputs, including fabricated quotes or citations (OpenAI), Anthropic says users should not rely on Claude as a singular source of truth (Anthropic), and Google advises double-checking Gemini responses and not using them for professional advice (Google).

A prompt to paste

Paste the rules of your task into the prompt, not the other way around. That is what keeps the tool inside the boundary your unit sets.

You are helping me with a university assessment. The AI rules for this task, copied from my unit outline or assessment brief, are:
[paste the exact wording of the AI rule for this assessment]

My task: [assignment question, word count, due date, subject]
Where AI is allowed for this task: [brainstorming only / outlining only / drafting with acknowledgement / not allowed at all]

Work within those rules and do this:
1. Before generating anything, tell me in one sentence what you think the question is asking.
2. Give me a section outline, and for each section list the evidence or sources I will need to find myself.
3. For each section, list the questions I still have to answer in my own words.
4. Do not write final paragraphs unless the rule I pasted allows AI drafting. If it does, draft only the section I name.
5. Flag anything you are not sure about instead of guessing. Do not invent sources, quotes, page numbers or statistics.
6. Finish with a short AI-use acknowledgement stating what you helped with, the tool and model, and the date, so I can check it against my unit outline.

Useful follow-ups:

Here are my rough notes: [paste your notes]. Do not rewrite them for me. Tell me what is missing, what does not answer the question, and what a marker would push back on.

Write one plain acknowledgement sentence I can adapt: I used [tool name] for [specific purpose] on [date]. Keep it factual and short, and tell me what my institution's preferred citation style might also require.

How far AI gets you

AI assists. The rules are set by your institution and your task, not by the tool, and Turnitin places the final decision with the instructor applying institutional policy — so AI can plan, explain, question and draft within the line your unit outline draws, but it cannot tell you where that line is or stand behind what it produces.

What you still do

Check the rule for each assessment. At the University of Sydney, your unit of study outline states whether AI tools can be used for an assessment, and breaching the guidelines can be an academic integrity breach (University of Sydney). TEQSA advises the same discipline-by-discipline check, because rules can differ between disciplines and even between tasks (TEQSA).
Treat "allowed" as a spectrum, not a yes. TEQSA says that depending on your institution's policies, using AI may be restricted or banned, and that some subjects or tasks may encourage or even require it (TEQSA). The University of Sydney allows AI tools as a general learning aid but limits their use in assessments (University of Sydney).
Acknowledge and cite permitted use. At the University of Melbourne, permitted AI use must be acknowledged and cited in accordance with the Assessment and Results Policy (MPF1326), and using AI where a task does not permit it — or without acknowledgement — is academic misconduct; students are responsible for checking the assessment guidelines (University of Melbourne). TEQSA also says that where AI use is permitted you need to understand how its use is referenced (TEQSA).
Keep your own evidence. The University of Sydney advises keeping track throughout the assignment process of all the ways you used generative AI, and keeping copies of generative AI outputs as evidence (University of Sydney). That record is what helps you if you are asked to explain your work (University of Melbourne).
Verify every fact yourself. Check each source, quote and figure the tool gives you against the original before it goes into your work.
Ask the official source when the rule is unclear. Your unit outline, unit coordinator, or your institution's academic integrity or student support office is the authority here, not a detector score and not a generic guide. If a concern or allegation is already in progress, use your institution's published academic integrity process and the support it offers students.

Common questions

Is using AI cheating?

It depends entirely on your task's rules. APA Style says submitting AI-generated work as though you did it yourself would be considered unethical — academic misconduct for students — and that what AI use is allowed, and how it should be disclosed, depends on instructor, departmental and publisher policies (APA Style). TEQSA warns that using AI inconsistently with your institution's rules can result in a finding of academic misconduct (TEQSA).

Can teachers detect AI, and is the score enough?

Instructors can see Turnitin's AI writing indicator; you cannot. But Turnitin says the score should not be the sole basis for action, and the University of Melbourne says a detection report alone is not sufficient evidence for an allegation — you may instead be asked to explain your work or provide drafts (Turnitin; University of Melbourne).

How do I cite AI use?

APA Style's 2025 template for a specific AI chat is: AI Company Name. (year, month day). Title of chat [Description, such as Generative AI chat]. Tool Name/Model. URL of the chat (APA Style). Your institution may require something different, so match the style your unit outline or style guide asks for (TEQSA).