Anthropic has introduced the AI Fluency Index, a new benchmark to measure how effectively people collaborate with AI based on real-world usage of its Claude models. The index aims to move beyond basic usage metrics to capture how skilfully individuals and organisations actually work with AI systems in practice.
Behavioural View of AI Fluency
The AI Fluency Index is grounded in a behavioural framework that defines key habits associated with safe and effective human–AI collaboration. In its first iteration, Anthropic focuses on 11 observable behaviours in Claude.ai and Claude Code chats, such as iterating on drafts, clarifying goals, questioning the model’s reasoning, and identifying missing context, to produce a baseline view of current AI fluency.
Iteration as the Strongest Signal
A central finding is that iteration – treating AI as a thought partner rather than a one-shot answer engine – is strongly linked to higher AI fluency. Conversations where users refine and build on Claude’s responses show far richer fluency behaviours overall, and those users are significantly more likely to question the system’s reasoning or flag gaps in its outputs.
Scrutiny Drops for Polished Outputs
The report also highlights a drop in scrutiny when Claude produces polished, “finished” outputs such as documents, code, or tools. In these cases, users tend to become more directive about goals, formats, and examples, but are less inclined to identify missing context, fact-check important claims, or unpack how the model arrived at its conclusions.
A Strategy Tool for AI literacy
Anthropic is positioning the AI Fluency Index as a strategic tool for educators and enterprises to diagnose gaps in AI literacy and shape targeted training programmes. By quantifying behaviours like iteration, goal setting, and evaluation, the company argues the index can reveal whether teams are merely delegating tasks to AI or building higher-skill, augmentative collaboration – and it plans to extend the work with studies comparing novice and expert users over time.
ALSO READ: Fei-Fei Li’s World Labs Raises $1 Billion to Scale Spatial AI
