Can You Trust the Text? Guide to AI Writing Detectors

As AI writing becomes indistinguishable from human writing, the need to detect it grows—especially in schools and newsrooms. But do these tools actually work?

1. GPTZero

One of the first and most popular detectors, GPTZero measures "perplexity" and "burstiness" to determine if text was written by a human (who tends to be more chaotic) or a machine (which is more predictable).

2. Originality.ai

Marketed towards publishers, Originality.ai claims to detect GTP-4 text with high accuracy. It's useful for website owners who want to ensure their freelance writers aren't just copy-pasting from ChatGPT.

3. The Accuracy Problem

No detector is 100% accurate. False positives (accusing a human of using AI) are a real problem, especially for non-native English speakers. These tools should be used as a signal, not as absolute proof.

Conclusion

The cat-and-mouse game between AI generators and AI detectors will continue. Ultimately, we may stop caring who wrote it, and start caring more about what it says.