Is AI humanizer text detectable by Turnitin now?
Short version of the mechanics: a detector is a classifier that scores how much a passage resembles text it was trained to call machine-written. A rewrite changes the surface features it leans on, so the score moves. An independent arXiv paper by Krishna and colleagues built a paraphrasing model and reported that it sharply lowered...
Cite
DigiCruncher, "Is AI humanizer text detectable by Turnitin now?", https://digicruncher.com/t/is-ai-humanizer-detectable-by-turnitin-now/302, accessed 2026-10-25.My current understanding, and I may be wrong: yes, Turnitin can often still flag text that went through an AI humanizer, because the company says it now looks for paraphrased and "bypassed" AI text as well as raw chatbot output. But I keep reading the opposite too, that a humanizer gets text past it, so I am not sure which is true today. I would rather know the limits than guess.
Some background. I am in a nursing program, and we write a lot of care plans and reflective essays. Last semester a classmate got an email about an "AI score" on a paper she says she wrote herself. It turned out fine after she showed her notes, but it scared me. Since then I use ChatGPT only to quiz myself on pharmacology and to explain things I do not follow, and I have never pasted its output into an assignment. Our syllabus says AI use must be disclosed, and one instructor said none at all on graded reflections.
What is bothering me is the group chat. Someone said a humanizer makes Turnitin unable to tell. Someone else said Turnitin updated this year and catches those now. A third person said the percentage is basically random. I read three articles and they contradicted each other, and a couple were written by companies that sell the humanizers, so I do not trust them.
What I would like to understand:
- What does Turnitin itself say it detects, as opposed to what blogs say?
- When a score shows up, is it a probability or a verdict, and how seriously do instructors take it?
- If I ever did use AI for an early draft with permission, what is the right way to do it so I am not at risk later?
Thanks, and sorry if this has been asked before.
1 more reply
Most helpful first- Ilya V.Member ·
Short version of the mechanics: a detector is a classifier that scores how much a passage resembles text it was trained to call machine-written. A rewrite changes the surface features it leans on, so the score moves. An independent arXiv paper by Krishna and colleagues built a paraphrasing model and reported that it sharply lowered detection for several detectors, including GPTZero and OpenAI's old classifier. Turnitin was not among those tested, so it says nothing direct about Turnitin. What it shows is the general shape: detectors and rewriters chase each other, so "it catches them" and "it cannot" both carry an expiry date.
Also, "basically random" is wrong. It is noisy, and Turnitin says it is noisiest at low percentages.
Write something first.
Give people something to work with: at least 30 words on what happened and what you tried.
That is too long. Keep it under 6,000 characters.
Write the question as the title, 15 to 140 characters, no links.
Pick a category.
Add a name (2 to 40 characters, no links).
That email address does not look right.
That was quick. Read the thread, then try again.
The form expired. Reload the page and post again.
Something went wrong with the form. Reload and try again.
Please complete the check and post again.
Limit reached for now. Try again later.
This thread is closed to new replies.
This thread no longer accepts replies.
Something went wrong with the form. Reload and try again.
Post a reply or question first (name and email), then this browser can vote, edit and accept answers.
You cannot vote on your own post.
Only the author (within 30 days) or the forum team can do that.
That email already has an account. Use your sign-in link instead.
New accounts are paused for the moment. Try again later.
That was already posted.