In March 2026, Anthropic launched a qualitative survey面向 Claude.ai users: How do you use AI? What do you hope it can achieve? What do you worry it might do? Nearly 81,000 people participated, covering multiple languages — the largest qualitative study of AI users to date. On April 28, Anthropic published the complete survey report.
Key Findings
Core Role Positioning of AI
Survey results show that user expectations for AI are shifting from “efficiency tool” to “thinking partner.” Top usage scenarios include:
- Writing and content creation: Report drafting, email composition, copy optimization
- Learning and research: Concept explanation, paper summaries, knowledge organization
- Programming assistance: Code generation, debugging, technical documentation
- Creative work: Brainstorming, solution design, visual prototyping
- Data analysis: Spreadsheet processing, trend identification, decision support
Notably, “creative work” ranked significantly higher than the same period last year, echoing Anthropic’s April 28 launch of “Claude for Creative Work” — user needs and product direction are converging.
What Users Expect from AI
Most respondents said they value three qualities in AI most:
- Contextual understanding: Remembering conversation history, understanding project background, not needing repeated explanations
- Output consistency and reliability: Fewer errors, fewer hallucinations, verifiable results
- Privacy and data security: Not wanting personal data used for training or leaked
The third point directly relates to Anthropic’s February announcement that “Claude will remain ad-free” — users’ sensitivity to commercialization models is rising.
User Concerns
The most集中 concerns in the survey include:
- Information accuracy: Factual errors and hallucinations in AI-generated content
- Job displacement: Impact of AI automation on specific roles
- Dependency: Over-reliance on AI leading to degradation of own abilities
- Data privacy: Where personal and corporate data flows and how it’s used
Implications for Product Selection
From the survey, it’s clear that user needs for AI products have become highly differentiated:
- Efficiency users (~40%): Value speed and accuracy — Claude Sonnet or GPT-5.4 level models are suitable
- Deep users (~30%): Need long context and deep reasoning — Claude Opus 4.7 or GPT-5.7 is a better match
- Creative users (~20%): Need multimodal and creative generation — Gemini 3.1 Pro or Claude Design is preferred
- Research/academic users (~10%): Need high accuracy and citability — requires specialized knowledge retrieval tools
Industry Trend Assessment
The sample size of 81,000 gives this survey high statistical significance. Several trends worth noting:
- AI is moving from “optional tool” to “daily infrastructure” — user frequency and scenario breadth are both increasing
- Multimodal demand is growing rapidly — pure text AI can no longer fully satisfy users; image, audio, and video processing capabilities are becoming new competitive points
- Privacy and data security have become core decision factors — users are giving increasing weight to privacy policies when choosing AI products
This survey provides AI practitioners with a rare user perspective: not “what is technically possible” but “what users actually need.” This data is worth referencing in product selection and roadmap planning.