Kush_EXE

How AI Detectors Work: Why One Essay Can Score 100% and 0%

· Runs 3:02

How AI detectors work: each one scores your finished text, not how you wrote it, with its own model and its own line, so one essay can get 100% AI from one detector and 0% from two others. We start with a real case: at Adelphi University a professor's report said Turnitin scored an essay 100%, the student said two detectors his parents ran showed 0%, and a New York judge annulled the violation in January 2026, over process and evidence, not detector accuracy.

From there we follow the idea: perplexity and burstiness, why GPTZero called part of the US Constitution likely AI in 2023, OpenAI's own classifier (26% of AI text caught in its test, 9% of human writing wrongly flagged, then pulled for low accuracy), and the deep learning models GPTZero and Turnitin describe now. Then we cover what a Turnitin percentage means, what recent tests found on false positives and missed AI, and what to bring if you're flagged: drafts, version history and your school's rule on tools like Grammarly.

Watch on YouTube Subscribe

Cheat sheet

An AI detector scores your finished text for patterns that resemble AI writing, not how you wrote it. The result is a statistical estimate. Turnitin and GPTZero say it is not a verdict.

What the detector actually measures

  • Predictability (early tools). Perplexity is how surprising text is to a language model. Low perplexity leaned AI. Ars Technica's example: "a cup of coffee" is low perplexity, "a cup of spiders" is high.
  • Burstiness. GPTZero defined it as how much writing patterns and perplexity vary across a document.
  • Learned patterns (now). GPTZero says it moved to deep learning in autumn 2023, though the same page still calls both one of seven indicators. Turnitin says its model is not explicitly programmed for either, but still explains detection through word probability.
  • A threshold. Turnitin scores each sentence from 0 to 1. The document percentage is the share over its threshold.

What the evidence says

MythEvidence
ChatGPT's maker has a working detectorOpenAI's classifier caught 26% of AI text and wrongly flagged 9% of human text in its 2023 test, then was withdrawn that July for low accuracy.
Detectors mostly flag human writingA 2025 working paper put GPTZero's false positive rate at about 0.7% and Pangram's near zero. Not peer reviewed, no student essays, and its authors expect an arms race.
Few false alarms means AI gets caughtIn a small, peer-reviewed 2026 test, Turnitin, GPTZero and Copyleaks failed on fully AI papers from ChatGPT Deep Research.
A 1% error rate is harmlessVanderbilt estimated Turnitin's claimed 1% would have meant about 750 of Vanderbilt's 75,000 papers from 2022 wrongly labelled: a projection, not a count.

What it cannot tell

  • Predictable human writing. In 2023, GPTZero said part of the US Constitution was likely written entirely by AI. People write predictably too, especially in formal styles such as law.
  • Non-native writers. In a 2023 Stanford study, seven detectors averaged about 61% false positives on 91 TOEFL essays. That finding is contested, and GPTZero self-reports 1.1% false positives on TOEFL texts.
  • Short or mixed text. Detection weakens sharply on short texts and joint human-and-AI essays.
  • The trade-off. Within one detector, fewer false positives means more missed AI text, and the reverse.

If your work is flagged

  • Turnitin says its percentage should not be the sole basis for action and does not determine misconduct.
  • Researchers say a flag should be paired with drafting history, so keep drafts and version history.
  • Detectors can disagree on the same essay, which you can raise in a dispute.

This cheat sheet as plain Markdown

In this video

  1. One essay, 100% and 0%: the Adelphi case
  2. Why detectors disagree, and the bet on predictability
  3. Perplexity and burstiness (a cup of spiders)
  4. Why the US Constitution got flagged as AI
  5. OpenAI's own classifier: 26% caught, then pulled
  6. From perplexity to deep learning
  7. What a 100% score means (Turnitin, GPTZero)
  8. False positives on human writing
  9. Where detectors miss AI text
  10. False alarms vs missed AI (Vanderbilt's 750)
  11. If you're flagged: what to bring

One essay, 100% and 0%: the Adelphi case 0:00

Your essay gets 100 percent AI from one detector, zero from two. That happened at Adelphi University: the professor's report said Turnitin scored it 100 percent, and accused the student of using Grammarly as generative AI. He denied writing it with AI, and said two detectors his parents ran showed zero. A New York judge annulled the violation in January 2026, over process and evidence, not detector accuracy.

Why detectors disagree, and the bet on predictability 0:26

How can one essay get 100 and zero? Each detector has its own model and line, and scores the finished text, not how you wrote it. Early detectors bet on predictability: AI picks the likely next word, and people are quirkier.

Perplexity and burstiness (a cup of spiders) 0:41

"I'd like a cup of coffee" surprises nobody. "A cup of spiders" does. That surprise is perplexity: low leaned AI. GPTZero's 2023 dashboard also scored burstiness, how much it varies.

Why the US Constitution got flagged as AI 0:55

In 2023, GPTZero said part of the US Constitution was likely written entirely by AI. GPTZero's creator, Edward Tian, called it "a text fed repeatedly into the training data." People write predictably too, especially in formal styles like law.

OpenAI's own classifier: 26% caught, then pulled 1:12

OpenAI's own trained classifier caught 26 percent of AI text in its 2023 test, and wrongly flagged 9 percent of human writing. By July 2023, it was gone, for low accuracy. So can a detector tell?

From perplexity to deep learning 1:26

GPTZero says it moved to deep learning in 2023, and Turnitin says its model isn't explicitly programmed for perplexity, though it still explains detection through word probability.

What a 100% score means (Turnitin, GPTZero) 1:37

Today, Turnitin's model learns patterns and scores each sentence from 0 to 1. Your percentage is the share over its line, so 100 doesn't mean certain. GPTZero's demo calls its number a probability, meaning how similar a text is to AI writing. Similar, not proven.

False positives on human writing 1:55

Fully human writing looks safer: in a 2025 working paper, GPTZero wrongly flagged about 0.7 percent, Pangram almost none. The catch: no student essays, no peer review, and its authors expect an arms race.

Where detectors miss AI text 2:11

Catching AI is the weak spot: in a small 2026 test, three tools got no fully AI paper right. On 40 mixed papers, scores landed in the true AI range zero to 37 times, depending on the tool.

False alarms vs missed AI (Vanderbilt's 750) 2:24

Detectors pick a line: fewer false alarms usually means more missed AI. Vanderbilt figured Turnitin's claimed 1 percent could have meant about 750 of its 75,000 papers from 2022 wrongly labeled. That was one concern when it switched the detector off in 2023.

If you're flagged: what to bring 2:43

If you're flagged, bring drafts, version history and your school's rule on tools like Grammarly, and point out detectors can disagree. Even Turnitin says its score shouldn't be the sole basis for action. A detector scores how your text looks, but your drafts show how you wrote it.

What the sources say

  • A New York court opinion dated January 28, 2026 says a professor at Adelphi University ran a student's essay through Turnitin, which gave it an AI score of 100%, and that the violation report accused the student of using Grammarly as generative AI. It also says two other detectors run by his parents showed 0%. New York State Law Reporting Bureau
  • In January 2026 a New York judge annulled the violation Adelphi University had recorded against a student whose essay Turnitin scored 100% AI, and ordered it expunged. The trial-level ruling turned on the university's process and the weak basis for the charge, not on a finding about detector accuracy. New York State Law Reporting Bureau
  • Perplexity measures how surprising a text is to a language model, and low perplexity makes a perplexity-based detector more likely to call the text AI-written. In Ars Technica's July 2023 example, a request for a cup of coffee, tea or water has low perplexity, while a request for a cup of spiders has high perplexity. Ars Technica
  • Ars Technica reported in July 2023 that GPTZero rated a section of the US Constitution as likely written entirely by AI. GPTZero's creator, Edward Tian, told Ars that the Constitution is fed repeatedly into the training data of many language models, so they are trained to generate similar text. Ars Technica
  • OpenAI reported in January 2023 that its own AI text classifier, on a challenge set of English texts, correctly identified 26% of AI-written text and wrongly labelled human-written text as AI-written 9% of the time. OpenAI withdrew the classifier as of July 20, 2023, citing its low rate of accuracy. OpenAI
  • GPTZero says it moved to a deep-learning architecture in autumn 2023. Turnitin says its model is not explicitly programmed to look at perplexity or burstiness, though its FAQ still explains detection through differences in word probability between AI and human writing. GPTZero, Turnitin
  • Turnitin says its model gives each sentence a score from 0 to 1, labels sentences above a threshold as AI, and reports the percentage above that threshold as the document's score. Turnitin also says the percentage should not be the sole basis for action and that it does not determine misconduct. Turnitin
  • A 2025 University of Chicago working paper by Brian Jabarian and Alex Imas tested AI detectors on 1,992 human-written passages from before 2020: Pangram had near-zero false positives and GPTZero about 0.7%. It is not peer reviewed, used no student essays, and its authors do not expect the results to hold over time. NBER, Chicago Booth Review
  • A peer-reviewed 2026 study found four detectors scored all 40 fully human papers below 20% AI, but Turnitin, GPTZero and Copyleaks got no fully AI paper right; those papers were written with ChatGPT Deep Research. Of 40 hybrid papers, GPTZero put 0 in the expected AI range, Copyleaks 12, Turnitin 24 and Pangram 37. International Journal for Educational Integrity
  • Vanderbilt University disabled Turnitin's AI detector in August 2023. It estimated that Turnitin's claimed 1% false positive rate, applied to the 75,000 papers Vanderbilt submitted in 2022, would have meant around 750 papers wrongly labelled, a projection and not a count of real cases. Vanderbilt University

Sources

  1. 2026 NY Slip Op 26021 (Supreme Court, Nassau County, January 28, 2026) New York State Law Reporting Bureau
  2. Why AI writing detectors don't work Ars Technica
  3. GLTR: Statistical Detection and Visualization of Generated Text arXiv
  4. Perplexity, burstiness, and statistical AI detection GPTZero
  5. GPTZero Technology GPTZero
  6. How to Use GPTZero | GPTZero Tutorial GPTZero on YouTube
  7. What is GPTZero? | Demo Video 2025 GPTZero on YouTube
  8. New AI classifier for indicating AI-written text OpenAI
  9. Turnitin's AI writing detection capabilities FAQs Turnitin
  10. How should I review the AI Writing report? Turnitin
  11. Understanding the false positive rate for sentences of our AI writing detection capability Turnitin
  12. Understanding the false positive rate for sentences within our AI writing detection Turnitin on YouTube
  13. Artificial Writing and Automated Detection NBER
  14. Do AI Detectors Work Well Enough to Trust? Chicago Booth Review
  15. Who wrote this? Evaluating the reliability of AI detection tools in higher education International Journal for Educational Integrity
  16. Guidance on AI Detection and Why We're Disabling Turnitin's AI Detector Vanderbilt University
  17. Detecting AI-Generated Essays in Writing Assessment: Responsible Use and Generalizability Across LLMs arXiv
  18. Why AI Detection Fails for Academic Integrity arXiv
  19. Who Wrote This? Turing, Total Variation, and the Mathematics of AI-Text Detection arXiv
  20. AI Detectors Don't Work. Here's What to Do Instead. MIT Sloan Teaching & Learning Technologies

More in AI detection explained