Human‑Guided AI Is the Creative Locus

Experimental evidence suggests that generative AI produces its best, most creative outputs when paired with human direction; unguided models perform poorly on visual creativity tests, while human‑guided models approach, but do not surpass, human artists. Different modalities matter: large language models excel at verbal divergent‑thinking tasks, but image models need human prompts, curation, or editing to generate novelty judged as creative. — This reframes policy and cultural questions from 'will AI replace artists?' to 'how should law, labor rules, and platforms allocate credit, control, and revenue when creativity is a human–AI hybrid?'