Readers can't identify watermarked AI text

Hacker News
Read full post
A quiz tested whether people can distinguish AI-generated text watermarked with SynthID-Text from unwatermarked text. Results showed participants performed no better than random guessing, indicating watermarks are not easily detectable by readers. This suggests watermarking does not noticeably alter AI-generated text quality.

More in LLM & Text Generation

DeepSeek V4.1 Flash now available on AI Gateway

Covered by 3 sources

Harvey raises $550M more to develop AI tools for legal teams

SiliconANGLE