As marketing professionals recognize the importance of understanding how humans interact with AI chatbots, recent empirical research illuminates real-world usage patterns across millions of conversations. This analysis synthesizes findings from three major studies examining ChatGPT, Claude, and Microsoft Copilot, revealing adoption trends, primary applications, and strategic implications for knowledge-intensive industries.
1. ChatGPT: Rapid Global Adoption Focused on Guidance, Information, and Writing Tasks
ChatGPT has achieved unprecedented adoption, reaching 700 million weekly users by July 2025 — approximately 10% of the global adult population. This adoption rate surpasses historical technologies like electricity or the internet. Initial users were predominantly male (80% with masculine names), but gender distribution has since balanced, now slightly favoring feminine names. Usage growth is accelerating in lower-income countries, with growth rates over 4x higher in low-income nations by May 2025.
Work vs. Non-Work Dynamics
Work-related interactions have grown steadily, but non-work usage has expanded more rapidly, rising from 53% to over 70% of total messages. This suggests significant value in non-professional contexts such as household tasks, potentially rivaling or exceeding productivity gains in formal employment.
Primary Applications
Approximately 80% of usage falls into three categories:
- Practical Guidance (personalized advice like tutoring or ideation)
- Seeking Information (queries on events, products, or recipes)
- Writing (drafting, editing, or summarizing text)
Writing predominates in work settings at 40%, with two-thirds of instances involving modifications to user-provided content rather than original generation.
Notable Observations
Computer programming accounts for only 4.2% of messages, far below common assumptions. Social or emotional uses are minimal (1.9% for relationships and reflection, 0.4% for role-playing). The focus instead centers on decision support: 49% involve asking for information, 40% doing tasks like output generation, and 11% expressing opinions.
Demographic and Occupational Insights
Higher-educated individuals in professional roles utilize ChatGPT more for work-related purposes. Mapping to O*NET activities indicates that 81% of work messages involve information processing or problem-solving, positioning AI as a tool for enhancing productivity in knowledge-based occupations.
For marketing professionals, these patterns underscore ChatGPT's utility in content ideation and audience research. The estimated consumer surplus — potentially $97 billion annually in the US — indicates untapped value in everyday AI integration.
2. Claude: Geographic Disparities, Increasing Automation, and Enterprise Deployment
Anthropic's analysis of over 1 million Claude.ai conversations alongside enterprise API data reveals evolving usage patterns from December 2024 to August 2025.
Temporal Shifts
Coding remains the largest category at 36%, but educational tasks increased from 9% to 12%, and scientific applications from 6% to 7%. Delegation of tasks has risen, with directive interactions (complete handoffs to AI) growing from 27% to 39%, indicating improved model reliability and reduced need for iterative fixes.
Geographic Variations
The Anthropic AI Usage Index (AUI) correlates strongly with income levels. High performers include Singapore (4.6x expected) and Canada (2.9x), while emerging markets like India (0.27x) and Nigeria (0.2x) show lower adoption.
Usage Diversity by Adoption Level
High-adoption regions exhibit varied applications across education, science, and business, whereas lower-adoption areas focus heavily on coding (over 50% in India). Advanced users favor augmentation (collaborative workflows), while others lean toward automation.
Enterprise Perspectives
API traffic representing programmatic business use is coding-intensive but 77% automation-oriented. Cost sensitivity appears low — higher-cost tasks see greater usage — suggesting that perceived value and capabilities drive deployment.
3. Microsoft Copilot: Distinguishing Assistance from Direct Performance
Microsoft's study of 200,000 anonymized Copilot conversations from 2024 differentiates "user goals" (tasks users seek help with) from "AI actions" (tasks the AI performs), mapping both to O*NET work activities.
Goals vs. Actions
Common user goals include information gathering, writing, and communication. AI actions often involve providing information, assistance, writing, teaching, or advising. In 40% of cases, these differ — for example, a user troubleshooting receives instructional guidance.
Performance Metrics
Information gathering and writing yield the highest user satisfaction and completion rates.
Occupational Impacts
Top-scoring categories include Sales, Computer and Mathematical, Office Support, Social Services, Arts/Media, Business/Finance, and Education. All sectors show potential, though physical roles experience narrower effects.
Implications for Marketing Agencies and Professionals
These studies demonstrate that AI chatbots primarily serve as tools for information processing, content creation, and decision support, with growing non-work applications. For agencies, this means prioritizing AI for audience insights, personalized campaigns, and efficiency gains in knowledge work. However, geographic and access disparities necessitate inclusive strategies to mitigate inequalities.
Sources
- Chatterji, A., Cunningham, T., Deming, D. J., Hitzig, Z., Ong, C., Shan, C. Y., & Wadman, K. (2025). How People Use ChatGPT. NBER Working Paper No. 34255.
- Appel, R., McCrory, P., Tamkin, A., Stern, M., McCain, M., & Neylon, T. (2025). Anthropic Economic Index report: Uneven geographic and enterprise AI adoption. Anthropic.
- Tomlinson, K., Jaffe, S., Wang, W., Counts, S., & Suri, S. (2025). Working with AI: Measuring the Applicability of Generative AI to Occupations. arXiv preprint arXiv:2507.07935.
