Developers’ Favorite AI Tools in 2025

AI Tool Usage by Developers: A Survey Analysis

Key Concepts:

• AI coding tools (GitHub Copilot, Cursor, Claude, Gemini)
• IDE (VS Code, JetBrains, PyCharm)
• Model usage trends (OpenAI, Gemini, Anthropic, Deepseek, Meta, Mistral)
• Vendor lock-in
• Model diversification
• Build vs. Buy vs. Hybrid approach to AI
• Open-source models

Survey Overview

Two surveys were analyzed: the Pragmatic Engineer 2025 survey (approximately 3,000 developers) and the Artificial Analysis survey (approximately 1,000 people). Both surveys focused on understanding which AI tools developers are using. The demographic distribution in both surveys was similar, with a majority of respondents in engineering and technical roles. The Artificial Analysis survey had a higher representation of large enterprises and individuals.

Pragmatic Engineer Survey Findings

• Respondent Profile: Primarily experienced developers (5+ years) working in backend development (16% in frontend/mobile).
• AI Tool Usage: A majority use one or more AI tools; 4% do not use AI tools due to lack of perceived usefulness, company policies, or ethical concerns.
• Most Popular AI Tool: GitHub Copilot is the most widely used, followed by ChatGPT, Cursor, Claude, and Gemini. This was surprising, as it contrasts with the perception from Reddit and Twitter.
• Usage Trends (2024 vs. 2025):
  • GitHub Copilot: Increasing overall usage, but lower mention rate than last year.
  • Cursor and Claude: Upward trend.
  • Gemini: Downward or static trend, despite model improvements.
• Company Size Correlation:
  • Medium to large companies: GitHub Copilot is more prevalent.
  • Startups: Cursor, Claude, or ChatGPT are more common. This suggests vendor lock-in plays a role in larger enterprises.
• Languages: TypeScript and Python are the most widely used, with Python being the most loved.
• Coding Tools: VS Code is the most widely used IDE, followed by JetBrains IDEs (PyCharm). Cursor is third.
• Disliked Tools: Jira, Microsoft Teams, and Confluence are disliked due to being slow, complex, buggy, and bloated.
• Gemini Anomaly: Gemini usage is consistent across company sizes (around 8%), raising questions about its relatively low adoption compared to its capabilities. Possible explanations include vendor lock-in with Google Workspace and usage by Android developers.

Artificial Analysis Survey Findings

• Model Usage: Gemini and OpenAI models are the most widely used, followed by Anthropic, Deepseek, and Meta's Llama.
• Model Usage Changes (Last Year): Gemini experienced a significant increase in usage (31% to almost 80%), while OpenAI remained relatively stable. Claude also gained ground.
• XAI Market Share: XAI saw a 31% increase in market share.
• API Providers: OpenAI is the top API provider, followed by Google and Microsoft. Grok surprisingly ranks higher than Amazon Bedrock.
• Inference Adoption: Google has seen better adoption for inference over the past year.
• AI in Production: 45% of respondents are using AI in production, while 50% are prototyping or exploring use cases.
• AI Application: The majority of users are in engineering and research and development.
• Model Market Share Changes: Google, XAI, and Deepseek have gained share, while Meta and Mistral are losing share.
• Model Diversification: Developers are using an average of almost five models, up from three in 2024. This reflects the need for specialized models for different applications (e.g., smaller models for latency-sensitive tasks, reasoning models for planning).
• Build vs. Buy: 32% of respondents favor building AI solutions, 27% buying, and 25% a hybrid approach. Enterprises likely prefer buying or hybrid approaches, while startups tend to build.
• Chinese Models: Organizations are increasingly open to using Chinese models if hosted outside of China.

Synthesis/Conclusion

The surveys reveal interesting trends in AI tool usage among developers. GitHub Copilot is the most widely used coding assistant, but the landscape is evolving rapidly. Gemini's adoption is a point of interest, with the Pragmatic Engineer survey showing static usage and the Artificial Analysis survey showing significant growth. Developers are diversifying their model usage, choosing different models for different tasks. The willingness to use Chinese models is also increasing. These findings highlight the dynamic nature of the AI tool market and the importance of understanding the specific needs and contexts of different development teams.