How People Use AI in Unexpected Ways
180For the past year, many have believed that artificial intelligence is mainly a tool to boost productivity—helping us write emails, code, or summarize large documents. But recent data suggests that how people really use AI might be quite different from these assumptions. A major study by OpenRouter looked at over 100 trillion tokens—basically billions of conversations and interactions with large language models like ChatGPT, Claude, and others—to uncover the truth about AI usage around the world.
Insights from a Global AI Usage Study
OpenRouter is a platform that connects requests to more than 300 AI models from over 60 providers, including OpenAI, Anthropic, and open-source options like Meta’s LLaMA. With over half of its usage coming from outside the United States and serving millions of developers globally, the platform offers a broad view of how AI is actually used across different regions and user groups. Importantly, the study analyzed metadata from billions of interactions without reading the actual conversation texts, ensuring user privacy while revealing behavioral trends.
One key finding is that open-source AI models are growing rapidly, making up about a third of total usage by late 2025. This surge often follows major model releases, showing how quickly the landscape is changing. The data paints a clear picture: AI isn’t just for business or coding—its real-world use cases are much more diverse and surprising.
Creative Use Cases Outnumber Productivity
The most unexpected discovery from the study is that over half of all open-source AI model activity isn’t for work tasks at all. Instead, users are engaging in roleplay, storytelling, and creative projects. That’s right—many people are using AI for fun, character-driven conversations, interactive fiction, and gaming scenarios. In fact, more than 50% of interactions with open-source models fall into this category, far surpassing the amount used for programming assistance.
This challenges the common belief that large language models are mostly tools for writing code, emails, or summaries. The data shows that a significant portion of AI interaction is about companionship and exploration. Users are treating these models as structured roleplaying engines, with around 60% of roleplay tokens related to gaming or creative writing. This hidden use case is changing how companies think about the potential of AI products.
It also highlights a shift in user behavior—people are not just seeking efficiency, but entertainment and social interaction through AI. This creative side of AI use remains largely invisible but is reshaping expectations and future development of these models.
Programming Continues to Grow Fast
While roleplay dominates open-source AI usage, programming-related queries are the fastest-growing category across all AI models. At the start of 2025, coding questions made up only 11% of total AI interactions. By the year’s end, that number had more than doubled to over 50%. This rapid growth shows how deeply AI is becoming part of software development and coding workflows.
Average prompt lengths for programming tasks also increased, indicating more complex and lengthy coding questions. This trend suggests that developers are increasingly relying on AI to assist with more sophisticated programming challenges, making AI an integral part of the coding process.
The rise in coding use reflects AI’s expanding role in tech and software industries, helping developers work faster and more efficiently. It also points to a future where AI tools are essential companions in building new software and tech solutions.
Overall, the study reveals that AI’s real-world use is broader and more varied than many assume. From creative storytelling to coding, people are finding many ways to engage with these models that go beyond simple productivity. These insights will likely influence how AI companies develop their products and how users choose to interact with these powerful tools.
What do you think?
It is nice to know your opinion. Leave a comment.