Generative AI is now embedded in digital media and production workflows, but audiences often have little way to know how much of a given output came from a machine. A new arXiv preprint introduces the Reported Engagement with AI Level (REAL) Rating, a framework designed for disclosing human-AI collaboration. The authors argue that existing provenance standards and disclosure methods are not enough to keep up with this shift.
The REAL Rating is proposed as a way to make AI involvement more transparent, giving the public a clearer signal about how AI was used in the content they encounter. Because this is a single preprint, there are no competing claims to compare; the contribution is the proposed framework itself.
The paper stops short of detailing the full rating scale in the abstract, but its central premise is that disclosure needs to be more structured and consistent. As generative AI continues to spread through everyday media, frameworks like this may become a practical part of how AI-assisted work is labeled.