AI Detector Accuracy Test: I Built a Python Script to Send the Same Essay to 5 Tools at Once

Short summary

The author built a Python script using aiohttp to send the same essay to four AI detectors (GPTZero, Originality.ai, Winston AI, ZeroGPT) simultaneously, motivated by wildly inconsistent scores between tools on identical text. The script tests three essay versions: raw GPT-4o output, hand-edited, and processed through a humanizer tool. Results highlight that AI detection tools produce significantly different scores on the same input, and even show variance when the same tool is run twice on identical text.

• •Python script sends identical text to 4 AI detectors concurrently using aiohttp
• •Same essay scored 82% on one detector and 31% on another with no edits
• •Tests raw GPT-4o output, hand-edited text, and humanizer-processed text across tools

Generated with AI, which can make mistakes.