How people actually use ChatGPT (www.lovelydata.cz)

A new OpenAI study finds that ChatGPT is now used far more for everyday personal tasks than for work: roughly 70% of prompts are non-work (vs. 30% work-related), a shift from near parity in June 2024 to a dominance of consumer use by June 2025. Total monthly messages ballooned from 451 million to 2.627 billion over that period, signaling broadening adoption. Rather than heavy technical or therapeutic roles, most interactions are “lighter-lift” activity—planning, drafting, brainstorming, learning—while coding made up just 4.2% of prompts, data analysis 0.4%, and mental-health style queries 1.9%. For the AI/ML community this reframes product and research priorities: many technical workflows have migrated off-chat to specialized API-driven tools (e.g., developer IDE integrations), explaining low coding counts in the ChatGPT corpus, and suggesting future growth in domain-specific, privacy-preserving small language models (SLMs). The findings imply AI’s biggest near-term impact is as a ubiquitous co-pilot in consumer contexts, not wholesale job replacement—so engineers and researchers should focus on consumer UX, efficient on-device models, safety/guardrails for everyday use, and metrics that capture real-world, non-enterprise value.