Turnitin is the most widely used AI detector in schools, colleges, and universities. It's built into learning platforms like Canvas, Blackboard, and Moodle, which makes it the default AI detector most students are up against.
Turnitin's AI detection has become much stronger. Text from ChatGPT, Claude, and Gemini is flagged quickly, and paraphrased AI writing is now detected too.
This guide covers:
- How Turnitin's AI detector works
- Turnitin's AI bypasser detection update
- How accurate Turnitin really is
- What doesn't work to bypass Turnitin
- What actually works, and which AI humanizers passed our Turnitin test
🎯 How Turnitin's AI Detector Works
Turnitin doesn't only check for plagiarism. It also includes an AI writing report, which instructors can see after a submission is scanned.
Turnitin analyzes the structure and language patterns of your writing. It looks for:
- Perplexity: how predictable your word choices are
- Burstiness: how much sentence length and structure vary
- Consistency: the uniform, robotic patterns common in AI writing
- Over-formality: the polished, academic tone AI models tend to produce
Even if your writing is accurate and grammatically clean, Turnitin can flag it if it reads like AI, especially essays with a uniform tone and flow.
The AI writing report gives your document an overall AI percentage. Many instructors treat anything above about 20% AI as worth a closer look. See our guide on what percentage of AI is acceptable on Turnitin for more detail.
🔄 Turnitin's AI Bypasser Detection
On August 27, 2025, Turnitin released one of its biggest AI detection updates. Earlier models mainly flagged text from ChatGPT, Gemini, or Claude. This update added AI bypasser detection, built specifically to catch text that was run through AI humanizers and similar tools.
Turnitin's AI writing report now highlights text in two ways:
- Cyan highlights: AI-generated text, including text that went through an AI bypasser
- Purple highlights: AI-generated text that was also paraphrased
Both are combined into the overall AI percentage. As a result, many basic humanizers and paraphrasers that used to work now get flagged.
🔍 How Accurate Is Turnitin?
Turnitin is effective at detecting unedited AI writing, but it isn't perfect:
- False positives: genuine human writing can be flagged as AI, especially formal, simple, or heavily polished writing. Research shows AI detectors flag non-native English writers far more often.
- Low-confidence scores: scores between 1% and 19% are shown with an asterisk (*) instead of a number, because Turnitin considers them less reliable.
Because of these limits, some schools have disabled or declined to use AI detectors. For a full breakdown, see is Turnitin accurate?
❌ What Doesn't Work to Bypass Turnitin
Many popular tricks don't work against Turnitin's current models:
- Paraphrasers like QuillBot, Wordtune, and Grammarly: they reword sentences but don't change the underlying AI patterns. Turnitin specifically detects paraphrased AI text.
- Swapping words: replacing a few words keeps the same predictable structure.
- Rewriting line by line: it takes hours and usually still leaves the AI patterns in place.
- ChatGPT prompts and "jailbreaks": asking ChatGPT to sound more human, or running text through it several times, produces a different version of AI writing that Turnitin can still detect.
- Random spacing, character swaps, and symbols: Turnitin's newer models are built to catch these tricks, and they make your writing look worse.
✅ What Actually Works: AI Humanizers
The most reliable way to bypass Turnitin is an AI humanizer. Paraphrasers and manual edits mostly change the words, but Turnitin flags the underlying patterns: predictable word choice, uniform sentences, and robotic structure. A good humanizer rewrites those patterns so the text reads like a person wrote it.
For school essays, a humanizer also needs to handle what Turnitin users care about most:
- Passing Turnitin's AI bypasser detection, not just older models
- Keeping an academic tone that fits the assignment
- Staying close to your word count, so a 1,000-word essay doesn't turn into 1,500
- Keeping your meaning, so your arguments and sources still line up
Before submitting, read the output and run it through an AI detector like the free Verva AI detector. Here's how the most popular humanizers performed against Turnitin.
⚖️ Which AI Humanizers Bypass Turnitin?
Passing Turnitin is only half the job. The rewritten text also has to keep your meaning and read like real writing. We ran the same AI-generated essay, about 290 words on Macbeth, through several popular AI humanizers and submitted each result to Turnitin.
| AI Humanizer | Bypasses Turnitin | Keeps Meaning | Readability | Output Length |
|---|---|---|---|---|
| Verva | ✅ Consistently | ✅ Yes | ✅ Natural | 301 words |
| Undetectable AI | ✅ Yes | ✅ Yes | ✅ Readable | 324 words |
| StealthGPT | ⚠️ Sometimes | ❌ Changes meaning | ⚠️ Adds unrelated content | 474 words |
| Walter Writes AI | ⚠️ Rarely | ⚠️ Partly | ❌ Unusual characters, broken paragraphs | 364 words |
1. Verva
Verva's AI humanizer passed Turnitin with 0% detected as AI while keeping the essay's meaning, structure, and formal tone. The output stayed close to the original length, which matters for assignments with a word count.
2. Undetectable AI
Undetectable AI also passed Turnitin with 0% detected as AI. The output was clear and readable, kept the main points of the essay, and stayed reasonably close to the original length. It's a strong option, especially for longer essays and documents.
3. StealthGPT
StealthGPT passed Turnitin in some of our tests, but the output often drifted from the original. It changed the meaning of several points and added sentences that weren't in the original essay, growing it from about 290 words to 474. For an assignment, those changes would need heavy editing.
4. Walter Writes AI
Walter Writes AI rarely passed Turnitin in our tests, and it took several attempts to get a passing result. When it did pass, the text included unusual inserted characters and oddly broken paragraphs, which made it hard to use as written.
The takeaway: several humanizers can get a 0% score, but only a few do it consistently while keeping your writing usable. Always read the output before you submit it.
📌 FAQ
Can Turnitin detect AI writing?
Yes. Turnitin detects AI-generated text from ChatGPT, Claude, Gemini, and other models, as well as paraphrased AI writing.
Can Turnitin detect QuillBot or Grammarly?
Yes. QuillBot and Grammarly mainly paraphrase, which leaves the underlying AI patterns in place, so Turnitin still flags the text.
Can Turnitin detect AI humanizers?
Turnitin's AI bypasser detection catches many basic humanizers. Stronger humanizers that fully restructure text, like Verva, still passed Turnitin in our tests.
What does the asterisk (*) mean on a Turnitin AI score?
It means the score is between 1% and 19%. Turnitin shows these low scores as an asterisk because it considers them less reliable.
What's the easiest way to bypass Turnitin?
Run your essay through an AI humanizer like Verva, which passed Turnitin with 0% AI in our tests while keeping the essay's meaning and length. Read the result, then check it with an AI detector before submitting.
