Reviewing the ChatGPT Usage Study: My Analysis Plan

Reviewing the ChatGPT Usage Study: My Analysis Plan

The NBER Working Paper 34255 "How People Use ChatGPT" represents the most significant empirical study of AI adoption we've seen. Aaron Chatterji and colleagues tracked real usage patterns from ChatGPT's November 2022 launch through July 2025, capturing the journey to 10% global adult adoption.

This isn't just another AI paper. It's the first large‑scale, longitudinal analysis of how humans actually interact with capable AI in the wild. The implications span economics, technology design, policy, and our understanding of human‑AI relationships.

Here's how I plan to break down this dense, important work into digestible insights.

Why This Study Deserves Deep Analysis

Most AI research focuses on capabilities: what models can do in controlled settings. This study flips the script: what do people actually do with AI when left to their own devices? The difference matters enormously.

The scale is unprecedented—billions of interactions across millions of users over 2.5 years. The behavioral data is real, not self‑reported. The global scope captures diverse adoption patterns across economic and cultural contexts.

But the paper is dense, technical, and packed with findings that deserve individual attention. A single blog post can't do justice to insights that will influence AI development, business strategy, and policy for years to come.

My Five‑Part Analysis Framework

I'm structuring the review around five core themes, each substantial enough for its own deep dive:

Part 1: Methodology and Research Foundation

Published: How People Use ChatGPT: Part 1 — Methodology and Scope

Focus: Understanding how the research was conducted and what makes it unique

Key Questions:

  • What makes this study different from previous AI adoption research?
  • How robust is the methodology? What are the limitations?
  • What can and can't we conclude from this approach?
  • How does this compare to classic technology adoption studies?

Why This Matters: Without understanding the methodology, we can't properly interpret the findings. The research design shapes every conclusion.

Part 2: Global Adoption Patterns and Demographics

Planned

Focus: Who adopted ChatGPT, where, and how fast

Key Questions:

  • What does the path to 10% global adoption tell us about AI diffusion?
  • Why are lower‑income countries showing higher growth rates?
  • How is the gender gap evolving, and what drives demographic differences?
  • What do geographic patterns reveal about infrastructure and access?

Why This Matters: Adoption patterns predict future AI market dynamics and reveal barriers to access that policy makers need to address.

Part 3: Usage Evolution—From Experiment to Habit

Planned

Focus: The dramatic shift from work to personal usage over time

Key Questions:

  • What caused non‑work usage to jump from 53% to over 70%?
  • How does professional usage vary by industry and role?
  • What does the work‑personal split reveal about AI's value proposition?
  • How do usage patterns mature as users gain experience?

Why This Matters: Understanding usage evolution helps predict where AI adoption goes next and how to design better AI products.

Part 4: Conversation Analysis—What People Actually Say to AI

Planned

Focus: The content and patterns of human‑AI dialogue

Key Questions:

  • Why do "Practical Guidance," "Information Seeking," and "Writing" dominate (80% of interactions)?
  • What makes writing tasks so central to work‑related AI usage?
  • How sophisticated are user interactions becoming over time?
  • What does conversation content reveal about AI capabilities vs. limitations?

Why This Matters: Conversation patterns reveal both human needs and AI strengths, guiding development priorities and use case design.

Part 5: Economic Impact and Strategic Implications

Planned

Focus: What this means for business, policy, and the future of work

Key Questions:

  • How do we measure AI's economic value in knowledge work?
  • What are the productivity implications of widespread AI adoption?
  • How should organizations prepare for AI‑native workflows?
  • What policy challenges does mass AI adoption create?

Why This Matters: The economic implications will shape business strategy, workforce development, and regulatory approaches for the next decade.

Cross‑Cutting Themes I'll Track

Several important themes span multiple parts of the analysis:

Digital Divide Evolution: How AI access patterns differ from traditional technology adoption, and what this means for global equity.

Human‑AI Interaction Maturation: How people learn to work with AI, and how this shapes both human behavior and AI development.

Platform Effects: What's unique to ChatGPT vs. generalizable to AI adoption broadly.

Measurement Challenges: How we quantify AI value, adoption success, and societal impact.

My Analysis Approach

For each post, I'll follow a consistent framework:

Context Setting: Why this aspect matters and how it connects to broader AI trends

Data Deep Dive: What the numbers actually show, with careful attention to methodology and limitations

Comparative Analysis: How these findings relate to other technology adoption patterns and AI research

Implications Mapping: What this means for different stakeholders—developers, businesses, policymakers, users

Future Questions: What we still don't know and what research should come next

What I'm Looking For

Beyond summarizing findings, I'm analyzing this study for:

Surprises: Results that challenge conventional wisdom about AI adoption Patterns: Trends that predict future AI development and usage Gaps: Important questions the study raises but doesn't answer Applications: How organizations can apply these insights practically Methodology Lessons: What this approach teaches us about studying AI in the wild

Why This Matters Now

We're at an inflection point in AI adoption. The period this study covers—November 2022 to July 2025—captures the transition from AI as experiment to AI as infrastructure. Understanding how this transition happened helps us navigate what comes next.

The study's findings will influence:

  • How companies design AI products and features
  • How organizations integrate AI into workflows
  • How policymakers regulate AI access and usage
  • How researchers study human‑AI interaction
  • How we prepare for broader AI adoption

Engagement and Discussion

Each post will include specific questions for discussion and areas where I'd value reader input. This isn't just analysis—it's the beginning of a conversation about what these findings mean for all of us working in and around AI.

I'm particularly interested in:

  • Practitioners' experiences that confirm or challenge the study's findings
  • International perspectives on adoption patterns and barriers
  • Industry‑specific insights about AI usage evolution
  • Policy implications that the study suggests but doesn't fully explore

Timeline and Next Steps

I'll publish one post per week, allowing time for each analysis to develop fully and for reader discussion between posts. The complete series should wrap up by the end of October.

Between posts, I'll be diving deeper into the paper's methodology, cross‑referencing findings with other research, and connecting with practitioners who can provide real‑world context for the academic findings.

Join the Analysis

This study deserves careful attention from anyone working in AI, technology strategy, or digital transformation. The findings will influence how we build, deploy, and regulate AI for years to come.

Follow along for the complete analysis, and don't hesitate to share your own insights and questions. The best analysis happens in dialogue, not isolation.

Next up: Part 1 examines the study's methodology and research approach. What makes this research unique, and what should we keep in mind as we interpret the findings?