Helping students and educators recognize AI-generated content
By the NJEA Technology Committee
As generative AI quickly evolves, distinguishing between authentic human work and increasingly sophisticated AI-generated content, including writing, images and video, is now a core educational challenge. Since current AI-detection software is unreliable and often produces false positives, both students and educators must develop strong critical media literacy skills. Research shows that AI-detection software is often unreliable and prone to false positives, making a human-centered approach—one that is focused on evaluating style, accuracy and authenticity—the most effective strategy.
Recognizing AI-generated text
AI writing often appears polished but lacks the individuality of human work. It typically includes repetitive or formulaic phrases and misses personal voice or class-specific insight. AI hallucination is the term used to describe AI-generated content that seems real but is false, misleading or made up. When writing feels generic or avoids referencing discussions, lessons or personal experiences, it may indicate the use of AI.
Recognizing AI images and deepfakes
AI-generated images frequently include distorted details such as extra fingers, inconsistent shadows, warped backgrounds or misshapen text. Deepfake videos may show unnatural movements, poor lip-syncing, or mismatched audio and lighting. Teaching students to use reverse image search and metadata tools strengthens their ability to verify authenticity. Educators can use sites such as detectfakes.kellogg.northwestern.edu from Northwestern University’s Kellogg School of Management to have students review AI-generated images and foster discussion about their ethical use and potential misuse.
Instructional approaches for educators
Instead of focusing on policing AI usage, experts advise redesigning learning experiences. Assignments that involve analysis, personal reflection, interviews, or referencing specific class content are harder for AI to mimic. Process-based assessments such as drafts, AI-use statements and class discussions encourage transparency and accountability.
These principles apply to teaching students and educators of all ages about artificial intelligence and critical content recognition.
- Start young: Introduce foundational AI ideas beginning in the early years.
- Teach the mechanics: Focus instruction on how AI works, not just on operational use.
- Foster skepticism: Cultivate critical thinking by encouraging students to question, verify and be skeptical of all AI outputs.
- Address ethics: Cover essential topics such as data privacy, deepfakes and the responsible use of AI tools.
- Support educators: Provide teacher training to ensure educators have the skills to integrate AI effectively and knowledgeably into their lessons.
- Be contextual: Tailor the use and discussion of AI to the specific subject or context.
To prepare students for a world where digital manipulation is standard, schools must shift from relying on detection tools to teaching verification, critical thinking and digital provenance skills. This is done by tracing the origin, authenticity and creation process of online content to determine whether it is real or manipulated. By teaching students to verify sources, check metadata and confirm the origins of digital media, educators help them navigate an online world where misinformation and AI-generated content are increasingly common. The new normal is to treat all online content as potentially altered until proven otherwise.
The NJEA Technology Committee
Sabina A. Ellis, chair, Essex County
David Ahn, Bergen County
Christopher Bowman, Burlington County
Patricia M. Martel, Camden County
Jonathan A. Gonzalez, Cumberland County
Salvatore A. Randazzo, Gloucester County
Daniel G. Abbadessa, Hudson County
Olive M. Giles , Mercer County
Deana Baumert, Monmouth County
Raymond A. Vikete , Morris County
Lori E. Lalama, NJREA
Melissa Krupp, Ocean County
William A. Krakower, Passaic County
Christopher J. Cook, Sussex County
Jasmine Y. Slowik, Warren County
Resources
Britannica Education: Spotting AI
MIT Sloan: AI detectors don’t work. Here’s what to do instead