AI Detector: How They Work and Why They Are Flawed
Exposing the mathematical biases behind AI detectors. Learn why these black-box systems penalize non-native speakers and how to protect genuine work from false positives.
An AI detector told you that your essay was written by ChatGPT.
The problem? You wrote it yourself.
You're not paranoid. You're not alone. And you're not wrong. AI detectors - the tools that claim to identify AI-generated text with high accuracy - have a documented false positive problem. They can flag human writing as machine-generated, especially from non-native English speakers, students, and technical writers.
This article explains how AI detectors work, why they fail so often, and how to protect genuine work if you are falsely flagged.
What Is an AI Detector?
An AI detector is a tool that claims to distinguish between human-written text and text generated by large language models (LLMs) like ChatGPT, Claude, or Gemini.
They're used by:
- Universities - to catch students submitting AI-generated essays
- Employers - to screen job applications and internal communications
- Publishers - to verify authorship of submitted articles
- SEO professionals - to check if content will be penalized by Google
- Individuals - to check their own writing out of curiosity or anxiety
The market for AI detection is huge, and search interest in "AI detector" is enormous. But the technology behind these tools has serious, well-documented limits.
How AI Detectors Work: The Two Signals
Despite different branding and interfaces, nearly all AI detectors rely on the same two statistical measures: perplexity and burstiness.
Signal 1: Perplexity
- Low perplexity = highly predictable word choices = likely AI
- High perplexity = unpredictable word choices = likely human
The flaw: This assumes humans write unpredictably. But formal writing, technical documentation, and non-native English are all highly predictable - and they get flagged as AI.
Perplexity measures how "surprised" a language model is by a piece of text. Technically, it's the exponential of the cross-entropy loss - but in plain English, it answers: how predictable is this text?
AI detectors feed your text into a reference language model and measure how often the model correctly predicts the next word. If the model predicts correctly most of the time, your text is "too predictable" and gets flagged as AI.
Signal 2: Burstiness
- Low burstiness = uniform sentences = likely AI
- High burstiness = varied sentences = likely human
The flaw: Many humans write uniformly. Students following a template, professionals writing reports, non-native speakers using safe structures - all get flagged as AI.
Burstiness measures variation in sentence structure and length. Humans write with rhythm - short sentences, long sentences, fragments, run-ons. AI tends to produce uniform sentence lengths.
AI detectors calculate the standard deviation of sentence lengths and structural patterns. Low variation gets flagged.
Why AI Detectors Fail So Often
The Non-Native Speaker Problem
Result: Human writing by non-native speakers is flagged as AI far more often than writing by native speakers. A widely cited Stanford study found that popular detectors misclassified a large share of essays written by non-native English speakers as AI-generated.
This is the most documented and most unfair failure mode.
Non-native English speakers tend to write with:
- Simpler vocabulary (they use words they're confident about)
- More uniform sentence structures (they stick to patterns they've learned)
- Fewer idioms and cultural references
These are exactly the signals AI detectors associate with AI-generated text.
This isn't a bug. It's a feature of how the detectors work - and it's deeply problematic.
The Academic Writing Problem
Academic writing is structured, formal, and predictable by design. You follow conventions: abstract, introduction, methodology, results, discussion, conclusion. You use formal vocabulary. You avoid contractions. You cite sources.
AI detectors see this predictability and flag it as AI.
If you run a handful of published, human-written abstracts through free detectors yourself, you may well see some of them flagged as AI-generated.
The Technical Writing Problem
Technical documentation uses standardized terminology, predictable sentence structures, and formal tone. API documentation, software manuals, and engineering reports are all flagged as AI at high rates because they're written to be clear and consistent - not creative and varied.
The Moving Target Problem
AI detectors are calibrated on output from specific models. When new models release, the detectors' training data can become stale. They either:
- Flag new AI output as human (false negative)
- Flag human writing as AI because the statistical baselines shifted (false positive)
AI detectors are playing whack-a-mole against rapidly evolving language models. They can't keep up.
Major AI Detectors Compared
GPTZero
Created by: Edward Tian How it works: Perplexity + burstiness analysis, plus newer classifier models Best for: Quick informal checks Use with care for: Academic writing, non-native speakers
GPTZero was one of the first widely known AI detectors and set the pattern for others. It offers free basic use and a clean interface. Like every detector, its published accuracy comes from test conditions that may not match the variety of real-world writing.
Turnitin AI Detection
Created by: Turnitin How it works: Proprietary model Best for: Universities and institutions (integrated with Turnitin's plagiarism checker) Use with care for: Non-native English speakers, technical writing
Turnitin is one of the most widely used AI detectors in academia. Its integration with plagiarism checking makes it convenient, and Turnitin itself acknowledges that false positives can happen, which is a serious concern for students.
Originality.ai
Created by: Originality.ai team How it works: Custom model trained on multiple LLM outputs Best for: SEO professionals and content marketers Less suited for: Academic use (not designed for education)
Originality.ai is built for content marketers and publishers. It is a paid tool aimed at agencies and site owners who want to screen content at scale.
Copyleaks
Created by: Copyleaks How it works: Combines plagiarism detection with AI detection Best for: Organizations screening content Less suited for: Casual individual checks
Copyleaks combines plagiarism and AI detection in one platform. It is aimed largely at businesses and institutions with compliance needs.
Can You Trust AI Detectors?
Short answer: Not on their own.
AI detectors are probabilistic tools, not definitive tests. They provide a likelihood score, not a fact. Treating them as ground truth is like treating a metal detector as proof of treasure.
Research on detectors, including the Stanford work on non-native speakers, keeps finding the same pattern: AI detection is not a solved problem. It may not be solvable with current approaches, which is why a detector score should never be the only evidence.
How to Protect Your Genuine Work
Even though detectors are flawed, you may still be judged by one. If AI use is restricted where you are submitting, the goal is not to beat the scanner. It is to be able to show that the work is yours.
1. Follow the Rules and Disclose
- Read your institution's or client's AI policy before you start
- If you used AI for brainstorming or editing, say so in a short note
- Do not use AI where it is prohibited
2. Keep Evidence of Your Process
- Write in a tool with version history (Google Docs, Word with tracking)
- Save outlines, notes and earlier drafts
- Keep your sources and research trail
3. Write Like Yourself
- Include specific details only you know (names, dates, what happened)
- Mix short and long sentences the way you naturally would
- Share your own opinions and reasoning
4. If You Are Flagged, Ask for Human Review
Ask which passages were flagged and what other evidence exists. Offer your version history and talk through how you wrote the piece.
A note on tools: rwrt is an iPhone app that rewrites text for tone and clarity. It makes no claims about AI detectors and is not built to evade them.
The Legal and Ethical Landscape
- Regulation: New AI rules such as the EU AI Act put more pressure on AI systems used in education to be transparent and fair
- Disputes: Some students have challenged false AI detection flags through appeals and, in some cases, in court
- Academic pushback: Some universities have paused or turned off AI detection features because of reliability concerns
The ethical question isn't whether AI detection is useful - it's whether the harm of false positives outweighs the benefit of catching AI-generated submissions.
FAQ
Are AI detectors accurate?
Not reliably. Independent research has documented false positives, especially for non-native speakers and formal writing. They're probabilistic tools, not definitive tests.
Why do AI detectors flag non-native speakers?
Non-native English writing tends to be more uniform and predictable - simpler vocabulary, consistent sentence structures. These are the same signals detectors associate with AI, leading to disproportionately high false positive rates.
Can I appeal an AI detection flag at my university?
Yes. Many universities now require human review before taking action on AI detection flags. Ask for the specific evidence and request a manual evaluation.
Does Google penalize AI-generated content?
Google's official stance is that AI-generated content is allowed if it's helpful and follows their quality guidelines. What matters is quality and usefulness, not how the content was produced.
Does rwrt bypass AI detectors?
No. rwrt rewrites text for tone and clarity using personas like Professional, Casual, Academic or Creative. It makes no claims about AI detectors and is not built to evade them. If AI use is restricted, follow the rules and keep your drafts to defend genuine work.
Say it the way you would.
rwrt rewrites your text as Professional, Casual, Academic, Creative or a persona you create. Free to start.