A recent analysis of AI image generation suggests that artificial intelligence may be more limited in creativity than many people assume. Researchers conducted an experiment where two popular AI models repeatedly generated images from one another over dozens of iterations — like playing a visual “telephone game.” Over time, instead of producing ever more unique or surprising visuals, the outputs tended to converge on a small set of familiar themes such as pastoral landscapes, rainy nightscapes, coastal beaches, and fancy interiors. This pattern indicates that AI often falls back on formulaic motifs instead of genuinely novel ideas when its prompts drift.
Experts describe this tendency as similar to “visual elevator music,” where the imagery feels familiar and uninspired rather than surprising or emotionally compelling. Even with thousands of iterations, the diversity of results clustered around about a dozen common visual tropes. This highlights a significant constraint in AI’s creative capacity: it excels at remixing and recombining patterns from its training data, but struggles to produce truly original content that steps outside learned conventions.
This finding resonates with broader discussions in the creative community about the limitations of AI creativity. While AI can generate content quickly and support creative workflows, its outputs often feel derivative or aesthetically predictable compared with human-made art, which emerges from lived experience, intention, and emotional nuance rather than pattern statistics. Many artists and designers maintain that human creativity retains a unique value that AI cannot replicate, particularly when it comes to conveying meaning, expression, and cultural context.
Overall, the experiment suggests that AI’s “creativity” may be more about efficient recombination than genuine invention. As useful as AI can be as a productivity tool — helping with drafts, ideation, or technical execution — it should not be conflated with the deep, original creativity that humans bring to art and design. This reinforces the idea that while AI can augment creative processes, it does not replace the human capacity for true originality.
