Detecting AI is complex, considering it is both possible for text written by students to show patterns of text written by AI, and vice versa. That said, there are 9 methods you can use to determine with a high degree of certainty whether text was generated by AI or a student.

9 Methods to Detect AI-Generated Student Work

Analyze Writing Patterns and Style

Look for:

• Unusually consistent grammar and vocabulary
• Lack of personal voice
• Abrupt changes in writing style within a single piece

Check Revision History

Here's how:

• Use Google Docs' Version History or extensions like Draftback to analyze essay submissions
• Large chunks of text pasted in at once could be a red flag, considering this may be from a tool like ChatGPT

Conduct Oral Examinations

Ask students to:

• Explain key concepts from their work
• Describe their research and writing process
• Elaborate on specific points in their assignment

Use Multiple AI Detection Tools

Remember:

• No single tool is 100% accurate
• Use results as part of a holistic assessment, not definitive proof
• Be aware of potential biases against non-native English speakers

Look for Inconsistencies in Knowledge

Watch for:

• Information beyond what was taught in class
• Errors in specific course-related content
• Inconsistencies with the student's previously demonstrated knowledge

Examine Source Citations

Verify:

• Existence of cited sources
• Relevance of citations to the topic
• Accuracy of quoted information

Often, AI models will "hallucinate" sources; this is a telltale sign of AI-generated work.

Assess Personal Voice and Experiences

Look for a lack of:

• Genuine personal anecdotes
• Emotional depth in writing
• Unique perspectives related to the student's background

If these are not currently part of your assignment, consider bringing them in.

Check for Current Events and Recent Information

Remember:

• Many AI models have knowledge cutoff dates
• Accurate references to very recent events suggest human input

Compare with In-Class Writing Samples

Maintain:

• A portfolio of in-class writing for each student
• Awareness of each student's typical writing style and capabilities

Automating the AI Detection Process

Blueye analyzes text using the most accurate AI detector available, compares it against student writing samples, examines sources, and more — all on autopilot.

Limitations of AI Detection Tools

While AI detection tools can be helpful, they're not infallible:

• False positives are common, especially for non-native English speakers
• Some tools struggle with certain writing styles or genres
• AI is rapidly evolving, potentially outpacing detection methods

To mitigate these issues:

• Use multiple detection tools for cross-verification
• Consider detector results in combination with other assessment methods
• Stay updated on the latest AI and detection technologies

Proactive Prevention Strategies

Prevention is often more effective than detection:

• Clearly communicate expectations regarding AI use in your syllabus
• Design assignments that require personal reflection or unique data sets
• Consider making more writing in class
• Brainstorm assessment methods that go beyond the traditional essay

Balancing Detection with Responsible AI Integration

While detecting misuse is crucial, it's becoming increasingly important to:

• Teach students about ethical AI use
• Demonstrate how AI can enhance learning when used appropriately
• Encourage critical thinking about AI's role in education and society

What to Do When You Suspect AI Use

1. Maintain open communication
2. Discuss concerns with the student privately
3. Involve administrators or parents as necessary
4. Provide opportunities for students to demonstrate understanding in alternative ways (depending on if this is the first instance of AI use)