Skip to Main Content

Generative Systems (ChatGPT, DALL-E, etc.): Detection

A collection of guides, resources, and information about the use of complex generative systems for use in research, writing, and other academic uses.

Detection Samples

In response to the growing proliferation and breadth of generative systems, there has been a surge of programs designed to detect if text has been human-generated or system-generated, and in some cases to what extent. The tests below were done using ZeroGPT, a program designed to identify portions of text that may have been generated rather than written. At times the program is rather astute in its detection, but in other instances it can be quite obviously incorrect. Take note of these samples:

Article Abstract (from “Robots and the Sacred in Science and Science-Fiction: Theleological Implications of Artificial Intelligence” by Robert M. Geraci):

 

 

“Fortress Investment Group Set to Acquire Vice Out of Bankruptcy” (New York Times article, June 22nd, 2023, as of 9:21 AM):

 

 

“The Tyger” by William Blake (as seen here):

 

 

Monologue, Romeo & Juliet by William Shakespeare, Act III, Scene III (as seen here):

 

Test 16: ChatGPT Response to Query: “Write a monologue in the style of Shakespeare mimicking the one spoken by Friar Laurence during ‘Romeo & Juliet,’ Act III, Scene III”:

 

ChatGPT Response to Query:

 

Full text of the United States Constitution (as seen on the web page of the National Constitution Center here):

 

 

Sample Generative System Detectors