Methodology

The data preparation and classification process involves several automated steps:

• Classification and Summarization: AI interactions that have been redacted of identifying information are classified as “work” and “non-work” related using an automated classifier, and then directed into unique pipelines. These conversations are then summarized, with each pipeline focusing strictly on the facets required for its downstream classification.
  
• Clustering: Using Observation Clustering and Taxonomy Organisation (OCTO), a bespoke clustering and hierarchical taxonomy assignment tool developed by Google DeepMind, these summaries are grouped into semantically similar clusters. The observations in the cluster are then summarized into a cluster label tailored to downstream classification needs.
  
• Taxonomy Mapping and Classification: Finally, these clusters are mapped to official statistical taxonomies: the Bureau of Labor Statistics (BLS) 2024 American Time Use Survey (ATUS) Activity Lexicon for non-work interactions, and BLS 2018 Standard Occupational Classification (SOC), further broken down into occupational titles and tasks from Occupational Information Network (O*NET) Database v30.2 for work interactions. We then augment the cluster dataset with additional bespoke classifiers unrelated to official statistical taxonomies. Details about these specific annotations is presented in ATLAS report sections that rely on them.
  
• Validation: The pipeline’s performance was evaluated using three distinct methods: measuring pipeline accuracy on synthetic data, inter-rater agreement evaluation with and without the AI ratings, as well as human approval of AI labels. For the former, we generated a synthetic ground-truth dataset seeded from the most granular tiers of the O*NET-SOC and ATUS taxonomies and calculated the rate with which our pipeline recovered the ground truth categories.

Gemini models play several essential roles throughout: summarizing conversations and clusters, extending the descriptions of taxonomy categories with additional detail, performing LLM classification for official taxonomy mappings and bespoke annotations, as well as creating synthetic datasets to validate performance.

To place our findings in context, it is helpful to note where ATLAS v1.0 differs in scope and design from the methodologies used by economic studies of AI use by Anthropic (Handa et al. 2025, Massenkoff et al, 2026) and OpenAI (e.g. Chatterji et al, 2025).

In terms of data sources, our dataset pools a larger variety and quantity of user interactions than have been previously explored together, spanning a standalone conversational AI app, an AI search experience, and a developer-facing API entry point. In terms of technical methodology, although we rely on a clustering-based taxonomy mapping process similar to Anthropic, our implementation differs considerably - for instance, by improving scalability to support a large sample size (15M), introducing recursive traversal of multiple nested statistical taxonomies (combining the SOC and O*NET classification into a single pipeline), leveraging LLM-assisted taxonomy category annotation to improve classification context, randomizing classifier options to mitigate well documented biases in LLM classification arising from option order (e.g. Pezeshkpour and Hruschka, 2023), and validating classifier performance with synthetic data.

Finally, in terms of the analytical scope of the dataset, we expand on the previously limited analysis of AI uses outside work by mapping AI interactions to an established statistical taxonomy for household time use (ATUS) and exploring multi-language adoption patterns. For work-related topics, we bring in new lenses from economic literature, such as the Autor-Thompson (2025) task expertise and routine/non-routine and manual/cognitive task characteristics.

Another innovation in ATLAS, to the best of our knowledge, is adjusting for Google app penetration differences between countries when we investigate cross-country AI adoption patterns.

1. Non-Negligible AI Use: The proportion of an occupation's tasks where AI is used by at least 25 individual users.
2. Intensive AI Use: The proportion of an occupation's tasks where AI is used by at least 100 individual users.
3. Full Automation: The estimated percentage of work interactions where users requested end-to-end task execution by the AI model without iterative human intervention.
4. Non-Routine Cognitive: The estimated share of work interactions categorized as analytical, creative, or specialized problem-solving tasks (e.g., information retrieval, learning, coding, strategic planning, creative design).
5. Major Occupation Group: High-level occupational categories (such as Business and Financial Operations Occupations) based on the US Bureau of Labor Statistics Standard Occupational Classification (SOC) 2-digit level code, representing broad sectors of the economy.
6. Minor Occupation Group: Intermediate-level occupational categories (such as Business Operations Specialists) based on SOC 4-digit level code.
7. Broad Occupation: Specific, detailed job classifications (such as Database and Network Administrators and Architects) based on SOC 5-digit level code.
8. Share of AI Usage per Occupation Group: The relative volume of conversational AI queries associated with each occupation group per capita.
9. Work vs. Non-Work Share: The estimated proportion of conversational AI interactions associated with workplace activities compared to personal life.
10. Activity Group, Activity, Sub-activity: Categorization of daily non-work AI queries mapped to the American Time Use Survey (ATUS). Activity Groups form the major overarching domains (Tier 1), Activities establish the intermediate categories (Tier 2), and Sub-activities (Tier 3) break them down into specific everyday tasks and household routines.
11. Conversational AI: In ATLAS v1.0, Conversational AI refers specifically to the aggregated user interactions from two consumer-facing Google surfaces: the Gemini App and Google AI Mode (AIM).
12. Differential Privacy: A mathematical privacy standard that injects calibrated statistical noise into aggregate data, protecting individual user privacy while preserving macro-level economic insights.