Students are expected to comply with their institution's academic integrity policies regarding the use of AI tools. If an instructor's assignment guidelines prohibit the use of AI-generated images, students are responsible for adhering to those guidelines.  

While no tool provides a definitive determination, the following visual cues may indicate that an image was AI-generated. Faculty may consider these indicators as part of a holistic assessment alongside other contextual evidence: 

• Anatomically or biologically implausible details: structures, textures, or organisms that do not match known specimens, such as incorrect cell shapes, symmetry errors, or non-existent species characteristics. 
• Unnatural background elements: blurred or illogical surroundings inconsistent with a home lab or field environment. 
• Inconsistent lighting or shadows: light sources that are physically impossible or shadows that do not match the apparent scene. 
• Absence of real-world artifacts: AI-generated images often lack realistic imperfections such as fingerprints, minor blurring, reflections, or environmental context. 
• Watermarks or visual artifacts: subtle repeating patterns or distortions characteristic of generative models. 
• Text or labels that are illegible or nonsensical: AI models frequently produce garbled text within images.  

A plethora of tools are available to detect AI-generated images, and can be evaluated and used by faculty to authenticate potentially fabricated images. Science Interactive does not endorse, guarantee, or take responsibility for the accuracy of any of these tools. Results should be treated as one data point among many, and not as conclusive proof of AI generation or authenticity.

Faculty who choose to use AI image detection tools should follow these best practices to ensure fair and appropriate use of results: 

• Use multiple tools: Run the image through at least two different detection tools and compare results. Inconsistent results across tools should reduce confidence in any single determination. 
• Document your process: Save screenshots of detection results, including the tool used, the date, and the confidence score returned. This documentation supports any subsequent academic integrity process. 
• Seek corroborating evidence: Detection tool results alone are insufficient grounds for an academic integrity allegation. Look for additional evidence such as implausible image content, inconsistencies with assignment requirements, or a student's inability to describe how they captured the image. 
• Engage your institution's academic integrity office: Before taking any formal action, consult with your institution's academic integrity office or equivalent body. They are equipped to advise on procedural requirements and student rights. 
• Have a conversation with the student: Ask the student to describe their process for producing the image. Genuine misunderstandings about AI tool policy do occur, and a direct conversation often provides important context. 
• Protect student privacy: Do not upload student submissions to third-party tools without confirming that your institution's data privacy policies and any applicable regulations (e.g., FERPA) permit this use.