Anthropic Education Report: The AI Fluency Index

As artificial intelligence (AI) becomes increasingly integrated into our daily routines, the rate of adoption is remarkable. However, the mere presence of AI tools does not provide insight into their impact. A critical question arises: as AI becomes a staple in everyday life, are individuals developing the necessary skills to utilize it effectively? This report aims to explore the development of “fluency” with AI technology over time.

Understanding AI Fluency

Previous Anthropic Education Reports have examined the use of Claude, our AI model, by university students and educators. Findings indicated that students leverage Claude for tasks such as report generation and lab result analysis, while educators utilize it for lesson planning and automating routine tasks. This report seeks to delve deeper into how individuals enhance their proficiency with AI tools.

The Importance of AI Fluency

AI fluency refers to the ability to effectively collaborate with AI systems. As AI tools become more prevalent, understanding how users interact with these systems is essential. The AI Fluency Index serves as a benchmark for assessing the development of fluency behaviors over time, providing insights into how users can improve their interactions with AI.

Measuring AI Fluency

To quantify AI fluency, we employ the 4D AI Fluency Framework, developed by Professors Rick Dakan and Joseph Feller in collaboration with Anthropic. This framework outlines 24 specific behaviors that exemplify effective human-AI collaboration. Of these, 11 behaviors are directly observable during interactions with Claude on Claude.ai or Claude Code.

Observable and Unobservable Behaviors

The 11 observable behaviors include actions such as questioning AI outputs and refining responses. The remaining 13 behaviors, which occur outside the chat interface, are more challenging to track but are equally important. Future research will leverage qualitative methods to assess these unobservable behaviors, as they are crucial for understanding AI fluency.

Data Collection and Analysis

For this study, we analyzed 9,830 anonymized conversations that took place over a seven-day period in January 2026. Our privacy-preserving analysis tool allowed us to measure the presence or absence of the 11 observable behaviors. We ensured the reliability of our sample by verifying consistency across various days and languages.

The AI Fluency Index

The result of our analysis is the AI Fluency Index, which provides a baseline measurement of how individuals currently collaborate with AI. This index serves as a foundation for tracking the evolution of fluency behaviors as AI models continue to advance.

Key Findings

Our initial findings reveal two primary patterns in the use of Claude:

• A strong correlation between AI fluency and the iterative refinement of conversations.
• Changes in user fluency behaviors when creating artifacts such as code or documents.

Iteration and Refinement

One of the most significant patterns observed is the relationship between iteration and refinement and other AI fluency behaviors. An impressive 85.7% of conversations in our sample demonstrated iteration and refinement, where users built upon previous exchanges to enhance their work.

Impact of Iteration on Fluency

Conversations characterized by iteration and refinement exhibited an average of 2.67 additional fluency behaviors, compared to a non-iterative rate of 1.33. Notably, these conversations were 5.6 times more likely to involve users questioning Claude’s reasoning and four times more likely to identify missing context.

Creating Artifacts

In 12.3% of conversations, users engaged in generating artifacts, including code, documents, and interactive tools. During these discussions, users displayed different collaboration behaviors with AI.

Behavioral Trends in Artifact Conversations

Conversations involving artifacts showed higher rates of behaviors related to “description” and “delegation.” Users were more likely to clarify goals, specify formats, and provide examples. However, this directiveness did not correlate with increased evaluation or discernment; in fact, users were less likely to identify missing context or question the model’s reasoning.

Interpreting the Results

The observed patterns raise important questions about user behavior. One possible explanation is that users may perceive polished outputs as complete, reducing the need for further questioning. Additionally, tasks involving artifacts may prioritize aesthetics or functionality over factual precision. Users might also evaluate outputs through other channels, such as testing code or sharing drafts with colleagues, rather than within the initial conversation.

The Need for Critical Evaluation

As AI models become more capable of producing polished outputs, the ability to critically evaluate these outputs becomes increasingly important. Users must develop skills to assess the quality and accuracy of AI-generated content, whether through direct conversation or other means.

Developing Your AI Fluency

AI fluency is a skill that can be developed over time. Based on our findings, we recommend three key areas for improvement:

1. Engage in Iteration and Refinement: Treat initial responses as starting points. Ask follow-up questions, challenge any inaccuracies, and refine your queries to enhance the quality of the interaction.
2. Question Outputs: Always assess the outputs generated by AI. Evaluate the reasoning behind the responses and identify any missing context to ensure the quality of the work.
3. Practice Collaboration: Work with AI as a partner. Use it to augment your capabilities rather than relying solely on its outputs. This collaborative approach fosters deeper understanding and skill development.

Frequently Asked Questions

Note: The development of AI fluency is crucial as AI tools become more integrated into various aspects of work and life. By understanding and improving these skills, individuals can maximize the benefits of AI technology.