What metadata would you use to indicate LLM generated text? (mastodon.social)

A discussion among developers highlights the need for effective metadata to clearly indicate text generated by large language models (LLMs), such as ChatGPT. Participants argue that traditional HTML attributes, like the 'lang' attribute, aren't suitable due to their established uses that assist screen readers and automated translation tools. Instead, a proposed solution is to use custom attributes, such as <p data-author="ChatGPT">. This method adds clarity without disrupting existing standards. The conversation also touches on the implications of accurately labeling AI-generated content. A logical taxonomy is suggested, differentiating between types of AI involvement—such as entirely LLM-written text, LLM-assisted content, and text that has undergone AI editing. This level of detail could help readers gauge the nature of the content they are consuming and could be particularly important for maintaining transparency in digital publishing, as well as supporting content creators and developers in properly attributing and utilizing AI tools.