Can Professors Detect ChatGPT? How They Spot It & Anti-Tips
You’ve written your assignment using ChatGPT. But as you go to upload, you can’t help but wonder: Can professors tell?
Particularly with AI tools continuing to advance, it’s a question that’s on more students' minds than ever.
In this piece, we’ll examine exactly how professors and universities detect AI-created code and text, the tools they use to do so, and offer some guidelines for responsibly utilizing ChatGPT in academic work. Let's dive in!
Can Professors Detect ChatGPT?
Professors may be able to identify ChatGPT content in student submissions; the extent to which content can be identified varies. Universities and teachers may have a range of strategies for recognizing AI text generators, including both high-tech and more traditional investigative processes.
AI Detection Tools
Professors most frequently use AI-detection tools. There are programs professors use to check essays for signs of AI-writing. Some popular ones are:
Turnitin: Previously used mainly for plagiarism, Turnitin has recently introduced an AI detection capability to catch text that has been produced by models such as ChatGPT. Turnitin compares user submissions to a vast collection of documents to identify matching content and other stylistic cues. Many course platforms include Turnitin (such as Canvas) as an embedded tool, so it’s a common way for teachers to spot ChatGPT text.
GPTZero, Originality.ai, and Other Third-Party Detectors: These tools are intended for recognizing AI-generated text. They scour the text for signs and patterns characteristic of AI text.
Grammarly, Copyscape, and Unicheck: Teachers can use Grammarly, Copyscape and Unicheck to analyze writing for plagiarism and AI influence. Copyscape and Unicheck check for copying and artificial intelligence language; Grammarly checks for style and errors.
There are, however, false positives – some writing that is not intentionally trying to deceive that is flagged by these tools – and they are not 100% accurate. As AI technology improves, detection methods are also continuously changing.
How Do AI Detection Tools Work?
The AI detection tools check many important factors in the text:
Perplexity: This quantifies how complex the language is. AI-written content has low perplexity because it tends to follow a predictable word use style. Human generated language is going to have higher perplexity.
Burstiness: The variance in sentence structure and length. Human writing often has a wider range of sentence structure and complexity, a more consistent source of length, while AI-written text lengths are a consistent source of variation.Text with more burstiness is likely human-written.
Manual Detection Methods
Additionally, professors can actively look for AI-generated work. If a student produces a piece of writing that is drastically different from any of their other work, for example if their writing is more formal, more organized, or lacking personal context, this might be the case. Teachers who recognize writing patterns of their students are most likely to catch shifts that call their attention to AI use.
Watermarks in AI-Generated Content
In another effort to facilitate detection, OpenAI, the manufacturer of ChatGPT, is investigating the use of watermarks on AI-written text. Watermarks would help teachers track content to its original location, providing a more straightforward way to address whether students are integrating ChatGPT content in their assignments. This system is not yet fully formed, but it has the potential to even more accurately determine copied ChatGPT text.
Can Professors Detect ChatGPT When It's Used for Coding?
It’s harder for teachers to detect code generated by Autocode’s ChatGPT than it is for them to detect other AI-written essays or papers. Most plagiarism detectors don’t check for code.
While plagiarism detection tools for code are limited, there are several signs professors may look for:
Beyond-Expected Complexity: If a student’s code is oddly advanced or using methods that they have not yet been taught, the professor will probably suspect AI assistance. For instance, new learners that turn in an elegant or highly-optimized solution that’s too advanced for the level they are at right can make a professor suspicious.
Similarity Between Submissions: If many students are submitting the same or very similar code, AI may be involved; this is especially true if all students are using the same (e.g., ChatGPT) tool. Professors may notice patterns or solutions that are the same and appeal.
How to Prevent AI Detection in Writing
If you want to reduce the likelihood of AI-generated content being flagged by universities, here are two practical steps you can take:
1. Manually Rewriting the Content
Once you've written using ChatGPT, rewrite some parts of the generated text in your own words. This will add aspects of your personal writing voice and will avoid the common repetitive formatting of an AI. Look at sentence formatting, vocabulary, and the overall idea to help localize it in your written work.
2. Use an AI Humanizer Tool
AI humanizers like EssayDone are tools that can help make AI text sound more like it was written by a human. AI humanizers adjust the tone, the structure, and phrasing to make them sound more like they come from a human writer.
3. Add Personal Insights
Generated AI content rarely contains personal experience or subjective opinions. Offering your own opinions, personal experiences, and relevant insights into your topic can help differentiate your work as being written by a human. It also makes your content more original and less likely to be identified as AI.
4. Proofread and Edit
Occasionally, AI-created writing can read as artificial and/or make abrupt logical leaps. Carefully read and correect your writing. Check that all the ideas are connected in a logical way. If there's any content that sounds overly-robotic or repeats too often, it may sound similar to a trap so consider revising.
5. Avoid Over-optimization
AI, especially ChatGPT, can generate text that seems too perfect or polished. If you want to create a more human-like written text, try not to be perfect. Write the way ChatGPT speaks: throw in some small imperfections or inconsistencies.
FAQ
Can Professors Detect ChatGPT?
Yes, Style inconsistencies and AI detection tools like Turnitin can help professors detect ChatGPT-generated work.
Can Schools Detect ChatGPT?
Yes, ChatGPT content can be identified if schools have AI detection tools, especially at well-funded or technically strong institutions.
Is Turnitin Available for Students?
Turnitin, like an LMS (Learning Management System) such as Canvas or Moodle. That said, the report by the AI in Turnitin is usually only visible to instructors and not students themselves.
Can Universities Detect ChatGPT Code?
Currently, it's hard to detect AI-generated code, but if it's too similar to past submissions, that's a red flag.
Can Google Classroom Detect ChatGPT?
Google Classroom does not detect ChatGPT, but teachers can utilize third-party tools.
How Do I Use ChatGPT Without Being Detected?
Use ChatGPT for brainstorming and outlines.
Rewrite generated text manually.
Use AI humanizers for more natural text.
The End
Ultimately, as professors and institutions become more adept at recognizing AIs used to generate content, knowing what tools they use can help you choose the best course of action when incorporating ChatGPT into your educational assignments.
Just remember to utilize AI with responsibility, and ensure the final product still represents your own work and knowledge; that’s the best way to get ahead without compromising your academic ethics.