Text Detection

Overview

Hive's AI-Generated Text Detection API takes a passage of text and determines whether or not that text is AI-generated. The model was trained on a vast dataset of both human-written and artificially generated text from many genres, with a focus on essays and school assignments. This API allows schools, universities, and other educational programs to quickly and effectively screen student work to detect the use of ChatGPT and other popular AI text generation engines. Digital platforms can also use this API to enact site-wide bans on AI-generated text content.

Response

The AI-Generated Text Detection API has one head. This head gives one ai_generated classification for all text inputs. The accompanying confidence score, which ranges from 0.0 - 1.0, indicates whether or not the text is predicted to be AI-generated.

In addition to calculating one classification for the text in its entirety, the model also splits the text into smaller chunks and determines a classification for each in order to highlight the specific sections of the text that are most likely to be AI-generated. Each of these chunks is 2048 characters long.

To see an annotated example of an API response object for this model, you can visit our API reference page.