A comprehensive study analyzing metadata from billions of AI interactions without accessing actual conversation text reveals a fundamental disconnect between corporate positioning and actual usage patterns: more than half of all open-source AI model usage is for roleplay and creative storytelling rather than productivity applications Cryptopolitan. While roleplay dominates open-source usage, programming has become the fastest-growing category across all AI models, exploding from 11% to over 50% of total usage by end of 2025 Cryptopolitan.
Open-source AI models have grown to capture approximately one-third of total usage by late 2025, with over 50% of open-source model interactions falling into roleplay and creative storytelling, dwarfing even programming assistance Cryptopolitan. Average prompt lengths for programming tasks have grown fourfold, from around 1,500 tokens to over 6,000, with some code-related requests exceeding 20,000 tokens roughly equivalent to feeding an entire codebase into an AI model for analysis Cryptopolitan.
Chinese AI models now account for approximately 30% of global usage nearly triple their 13% share at the start of 2025, with DeepSeek alone processing 14.37 trillion tokens during the study period Cryptopolitan. Asia's overall share of AI spending more than doubled from 13% to 31%, with Singapore emerging as the second-largest country by usage after the United States Cryptopolitan.
The study introduces agentic inference as the defining concept of AI's next phase: AI models are no longer just answering single questions they're executing multi-step tasks, calling external tools, and reasoning across extended conversations, with the share of AI interactions classified as "reasoning-optimised" jumping from nearly zero in early 2025 to over 50% by year's end Cryptopolitan.
This reflects a fundamental shift from AI as a text generator to AI as an autonomous agent capable of planning and execution, with the median LLM request no longer a simple question or isolated instruction, but instead part of a structured, agent-like loop, invoking external tools, reasoning over state, and persisting across longer contexts Cryptopolitan. Simplified Chinese is now the second-most common language for AI interactions globally at 5% of total usage, behind only English at 83% Cryptopolitan, demonstrating rapid diversification beyond Western markets.
Researchers discovered the "Glass Slipper" effect a phenomenon where AI models that are "first to solve" a critical problem create lasting user loyalty Cryptopolitan, suggesting that competitive differentiation increasingly depends on being the first solution to address specific user pain points rather than achieving marginal performance improvements across benchmark tests.
The finding that over 50% of open-source interactions are roleplay and creative storytelling counters an assumption that LLMs are mostly used for writing code, emails, or summaries Cryptopolitan, according to the report. The data reveals that developers aren't just asking for simple code snippets anymore but conducting sophisticated debugging sessions, architectural reviews, and multi-step problem solving.
The rapid adoption of Chinese AI models including DeepSeek, Qwen from Alibaba, and Moonshot AI represents a fundamental shift in the global AI landscape, where Western companies no longer hold unchallenged dominance in model deployment and usage patterns.
For technology executives, the data suggests enterprise AI strategies may be misaligned with actual user behavior patterns, with significant portions of computational resources directed toward non-productivity applications. The shift toward reasoning-optimised interactions and agentic workflows means organizations must prepare infrastructure for multi-step autonomous operations rather than simple query-response patterns Cryptopolitan.
The fourfold increase in average programming prompt length indicates developers are moving beyond basic code generation toward comprehensive codebase analysis and architectural problem-solving, requiring substantially higher computational capacity and more sophisticated context management. Organizations deploying AI development tools must provision infrastructure capable of processing 20,000+ token requests routinely rather than treating them as edge cases, with Chinese model proliferation introducing geopolitical considerations into AI procurement and data sovereignty strategies.
The transition from AI as text generator to autonomous agent capable of planning and execution represents the defining trajectory for 2026 Cryptopolitan, with reasoning-optimised interactions expected to dominate usage patterns. Decision-makers should monitor whether the Glass Slipper effect creates winner-take-all dynamics in specific vertical applications, potentially fragmenting the AI market around specialized problem-solving capabilities rather than general-purpose models. The surprise dominance of roleplay and creative applications in open-source models suggests consumer AI adoption drivers differ fundamentally from enterprise priorities, requiring bifurcated product and infrastructure strategies.
Source & Date
Source: Artificial Analysis Study, AI News
Date: December 9, 2025
