How folks actually use AI: The stunning reality from analysing billions of interactions

For the previous 12 months, we’ve been informed that synthetic intelligence is revolutionising productiveness—serving to us write emails, generate code, and summarise paperwork. However what if the truth of how folks really use AI is totally completely different from what we’ve been led to consider?

An information-driven study by OpenRouter has simply pulled again the curtain on real-world AI utilization by analysing over 100 trillion tokens—primarily billions upon billions of conversations and interactions with massive language fashions like ChatGPT, Claude, and dozens of others. The findings problem many assumptions in regards to the AI revolution.

​​OpenRouter is a multi-model AI inference platform that routes requests throughout greater than 300 fashions from over 60 suppliers—from OpenAI and Anthropic to open-source alternate options like DeepSeek and Meta’s LLaMA. 

With over 50% of its utilization originating outdoors america and serving hundreds of thousands of builders globally, the platform provides a novel cross-section of how AI is definitely deployed throughout completely different geographies, use circumstances, and consumer sorts. 

Importantly, the research analysed metadata from billions of interactions with out accessing the precise textual content of conversations, preserving consumer privateness whereas revealing behavioural patterns.

The roleplay revolution no person noticed coming

Maybe essentially the most stunning discovery: greater than half of all open-source AI mannequin utilization isn’t for productiveness in any respect. It’s for roleplay and artistic storytelling.

Sure, you learn that proper. Whereas tech executives tout AI’s potential to rework enterprise, customers are spending the vast majority of their time partaking in character-driven conversations, interactive fiction, and gaming situations. 

Over 50% of open-source mannequin interactions fall into this class, dwarfing even programming help.

“This counters an assumption that LLMs are largely used for writing code, emails, or summaries,” the report states. “In actuality, many customers have interaction with these fashions for companionship or exploration.”

This isn’t simply informal chatting. The information exhibits customers deal with AI fashions as structured roleplaying engines, with 60% of roleplay tokens falling beneath particular gaming situations and artistic writing contexts. It’s an enormous, largely invisible use case that’s reshaping how AI corporations take into consideration their merchandise.

Programming’s meteoric rise

Whereas roleplay dominates open-source utilization, programming has change into the fastest-growing class throughout all AI fashions. At first of 2025, coding-related queries accounted for simply 11% of complete AI utilization. By the tip of the 12 months, that determine had exploded to over 50%.

This development displays AI’s deepening integration into software program improvement. Common immediate lengths for programming duties have grown fourfold, from round 1,500 tokens to over 6,000, with some code-related requests exceeding 20,000 tokens—roughly equal to feeding a whole codebase into an AI mannequin for evaluation.

For context, programming queries now generate a few of the longest and most advanced interactions in the complete AI ecosystem. Builders aren’t simply asking for easy code snippets anymore; they’re conducting refined debugging periods, architectural critiques, and multi-step downside fixing.

Anthropic’s Claude fashions dominate this area, capturing over 60% of programming-related utilization for many of 2025, although competitors is intensifying as Google, OpenAI, and open-source alternate options acquire floor.

The Chinese language AI surge

One other main revelation: Chinese language AI fashions now account for roughly 30% of world utilization—almost triple their 13% share in the beginning of 2025.

Fashions from DeepSeek, Qwen (Alibaba), and Moonshot AI have quickly gained traction, with DeepSeek alone processing 14.37 trillion tokens throughout the research interval. This represents a basic shift within the world AI panorama, the place Western corporations not maintain unchallenged dominance.

Simplified Chinese language is now the second-most widespread language for AI interactions globally at 5% of complete utilization, behind solely English at 83%. Asia’s total share of AI spending greater than doubled from 13% to 31%, with Singapore rising because the second-largest nation by utilization after america.

The rise of “Agentic” AI

The research introduces an idea that may outline AI’s subsequent section: agentic inference. This implies AI fashions are not simply answering single questions—they’re executing multi-step duties, calling exterior instruments, and reasoning throughout prolonged conversations.

The share of AI interactions categorized as “reasoning-optimised” jumped from almost zero in early 2025 to over 50% by 12 months’s finish. This displays a basic shift from AI as a textual content generator to AI as an autonomous agent able to planning and execution.

“The median LLM request is not a easy query or remoted instruction,” the researchers clarify. “As an alternative, it’s a part of a structured, agent-like loop, invoking exterior instruments, reasoning over state, and persisting throughout longer contexts.”

Consider it this manner: as a substitute of asking AI to “write a operate,” you’re now asking it to “debug this codebase, establish the efficiency bottleneck, and implement an answer”—and it will possibly really do it.

The “Glass Slipper Impact”

One of many research’s most fascinating insights pertains to consumer retention. Researchers found what they name the Cinderella “Glass Slipper” impact—a phenomenon the place AI fashions which might be “first to resolve” a essential downside create lasting consumer loyalty.

When a newly launched mannequin completely matches a beforehand unmet want—the metaphorical “glass slipper”—these early customers stick round far longer than later adopters. For instance, the June 2025 cohort of Google’s Gemini 2.5 Professional retained roughly 40% of customers at month 5, considerably increased than later cohorts.

This challenges standard knowledge about AI competitors. Being first issues, however particularly being first to resolve a high-value downside creates a sturdy aggressive benefit. Customers embed these fashions into their workflows, making switching pricey each technically and behaviorally.

Price doesn’t matter (as a lot as you’d assume)

Maybe counterintuitively, the research reveals that AI utilization is comparatively price-inelastic. A ten% lower in worth corresponds to solely a couple of 0.5-0.7% enhance in utilization.

Premium fashions from Anthropic and OpenAI command $2-35 per million tokens whereas sustaining excessive utilization, whereas funds choices like DeepSeek and Google’s Gemini Flash obtain related scale at beneath $0.40 per million tokens. Each coexist efficiently.

“The LLM market doesn’t appear to behave like a commodity simply but,” the report concludes. “Customers stability price with reasoning high quality, reliability, and breadth of functionality.”

This implies AI hasn’t change into a race to the underside on pricing. High quality, reliability, and functionality nonetheless command premiums—at the least for now.

What this implies going ahead

The OpenRouter research paints an image of real-world AI utilization that’s much more nuanced than business narratives recommend. Sure, AI is remodeling programming {and professional} work. But it surely’s additionally creating fully new classes of human-computer interplay via roleplay and artistic purposes.

The market is diversifying geographically, with China rising as a significant drive. The expertise is evolving from easy textual content technology to advanced, multi-step reasoning. And consumer loyalty relies upon much less on being first to market than on being first to really resolve an issue.

Because the report notes, “methods through which folks use LLMs don’t at all times align with expectations and range considerably nation by nation, state by state, use case by use case.”

Understanding these real-world patterns—not simply benchmark scores or advertising and marketing claims—shall be essential as AI turns into additional embedded in day by day life. The hole between how we expect AI is used and the way it’s really used is wider than most realise. This research helps shut that hole.

See additionally: Deep Cogito v2: Open-source AI that hones its reasoning skills

Wish to be taught extra about AI and large information from business leaders? Try AI & Big Data Expo happening in Amsterdam, California, and London. The great occasion is a part of TechEx and is co-located with different main expertise occasions together with the Cyber Security Expo. Click on here for extra info.

AI Information is powered by TechForge Media. Discover different upcoming enterprise expertise occasions and webinars here.