AI detector vs plagiarism checker: what's the difference?

Updated October 31, 2026 · 4 min read

Short answer

A plagiarism checker compares your text with existing sources (web pages, published papers and, in Turnitin's case, past student submissions) and shows the passages that match. An AI detector reads the writing itself, its word choice, structure and predictability, and estimates whether a language model wrote it, without comparing it with any database.

They answer different questions, so one can't stand in for the other. AI-generated text is new text, so it usually passes plagiarism checks: in a 2023 test, 25 ChatGPT essays averaged 8.76% similarity in iThenticate. Turnitin says its similarity score and its AI score are "completely independent and do not influence each other."

What does each tool check?

	Plagiarism checker	AI detector
The question it answers	Does this text match something that already exists?	Was this text likely written by an AI model?
How it works	Compares the text with a database of web pages, publications and, for some tools, past submissions	A trained model reads word choice, structure and predictability; no database
What you get	Matching passages with their sources, and a similarity percentage	An AI probability or share, often with the flagged passages
What a high score means	A lot of the text matches sources, which may still be properly quoted and cited	The writing looks statistically like AI output
What it misses	New text, including most AI text; contract cheating; heavily reworded copying	Copied human text; short, heavily edited or humanized AI text

Scribbr puts it simply: AI detectors work "by measuring specific characteristics of the text like sentence structure and length, word choice, and predictability — not by comparing it to a database of content," while plagiarism checkers compare "the text to a large database of web pages, news articles, journals, and so on." As QuillBot notes, "A piece of writing can be original but AI-generated, or copied from another source without involving AI."

Neither tool makes the final call. Turnitin says it "does not check for plagiarism in writing." It checks submissions against its database and highlights matches "for you to review." Whether a match is plagiarism or a properly cited quote is a human judgment.

Can a plagiarism checker detect ChatGPT?

Usually not. A language model writes new sentences rather than copying existing ones, so there's little for a plagiarism checker to match.

ChatGPT essays (Khalil and Er, 2023): 25 essays run through iThenticate averaged 8.76% similarity, and 68% scored under 10%. Another 25 run through Turnitin averaged 13.72%. It's a small test of ChatGPT as of January 2023.
Medical abstracts (Gao et al., npj Digital Medicine, 2023): abstracts written by ChatGPT had a median plagiarism score of 0% on a free plagiarism website. An AI-output detector separated them from real abstracts well (AUROC 0.94, where 1.0 is perfect). The authors' summary: "The generated abstracts do not always alarm plagiarism-detection models, as the text is generated anew, but can often be detected using AI detection models."
Turnitin's own warning: "a low score does not guarantee the absence of plagiarism." And a 0% similarity score "may indicate a document consisting entirely of original material" only "If there are no Integrity Flags, and no AI writing detected."

Can an AI detector find plagiarism?

No. An AI detector doesn't look anything up. Pangram, for example, says its detector "does not compare the text to other submissions in a database or on the web like traditional plagiarism checkers." Text copied word for word from a human-written web page would be scored as human-written, because it is.

That's why the two scores are reported separately. Turnitin says its similarity score and AI writing percentage "are completely independent and do not influence each other," and Copyleaks describes its AI percentage as separate from its similarity score too.

What happens with paraphrased text?

Paraphrasing tools can beat plagiarism checkers. In a 2022 study, the commercial checker PlagScan performed "close to random" on machine-paraphrased student theses, and people identified GPT-3 paraphrases with only 53% accuracy (Wahle et al., EMNLP 2022). Turnitin also warns that "content spinners" can produce low similarity scores for unoriginal work.
Paraphrasing also hurts AI detectors. In a 2023 test of 14 tools, AI text run through QuillBot was identified only 26% of the time (Weber-Wulff et al.). Detectors have changed since, but rewriting is still their weak spot. What newer research shows.
Vendors say they're catching up. Turnitin added detection of AI paraphrasing tools in December 2023 and of "AI bypasser" (humanizer) tools in August 2025 (that feature is English only). Copyleaks says its plagiarism checker detects "paraphrasing, image-based plagiarism, character manipulation, and even LLM-generated text." These are vendor claims: we found no independent test of Turnitin's bypasser detection, and Copyleaks publishes no method for its claim.

What is Turnitin Authorship?

A third kind of check. Turnitin says Authorship "uses metadata as well as forensic language analysis to detect if the submitted assignment was written by someone other than the student." It looks for ghostwriting and contract cheating, and Turnitin says it "will not be able to indicate if it was AI written."

Which check do you need?

Your situation	What to run
Checking your own essay before you submit it	Both: a plagiarism check catches quotes you forgot to cite, an AI check shows passages that read like AI. How to check your essay
Grading an essay that seems off	Both, plus drafts, sources and a conversation with the student. A fair process for teachers
Reviewing work from a freelance writer	Both, and agree on AI rules before the work starts. AI detectors and freelance writing
Submitting a thesis	Your graduate school's own similarity check, which may or may not include AI detection. How theses are checked
Finding out whether text was copied from a website	A plagiarism checker. An AI detector can't tell you.

AI Checker Plus runs both. The AI check labels each passage AI-generated, AI-assisted or human-written. The plagiarism check, on paid plans, matches the text against the web and lists the share of sentences copied, with links to each source.

Questions people also ask

Does Turnitin's similarity score include AI writing?

No. Turnitin says the similarity score and the AI writing percentage are completely independent and do not influence each other. The similarity score is the share of text that matches its database; the AI percentage is its model's estimate of how much text was generated or modified by AI.

Can ChatGPT text pass a plagiarism check?

Usually, because ChatGPT writes new sentences instead of copying existing ones. In a 2023 test, 25 ChatGPT essays averaged 8.76% similarity in iThenticate, and 68% scored under 10%. An AI detector is the tool built to answer that question.

Is a 0% similarity score proof that a text is original?

Not by itself. Turnitin says a 0% score may indicate fully original material only if there are no integrity flags and no AI writing detected. It also warns that contract cheating and content spinners can produce low similarity scores for work that isn't original.

Do I need both an AI check and a plagiarism check?

If you need to know both whether text was copied and whether it was written by AI, yes: each tool is blind to what the other looks for. Many services now offer both, including Turnitin (as separate reports), Copyleaks and AI Checker Plus.

Sources

By the AI Checker Plus team. We research and draft these guides with help from AI tools, then check every claim against the sources listed above.