The state of AI coding in 2025: Adoption, proficiency, and transformation
Most developers now use AI coding tools regularly, but proficiency levels vary widely. Our survey of 200 developers reveals how AI is reshaping software development workflows.
By Mihail EricKey findings
AI adoption is now nearly universal among developers: Eighty-nine percent of surveyed developers report using AI coding tools regularly—up from 64 percent just one year ago.
Proficiency remains a significant gap: While adoption is high, only 42 percent of developers consider themselves intermediate or advanced users of AI coding tools. Most are still learning to use these tools effectively.
Productivity gains are measurable but uneven: Developers report an average 3.2x increase in code completion speed, but benefits vary significantly based on use case and proficiency level.
Code review remains essential: Sixty-seven percent of developers manually review all AI-generated code before committing. Trust in AI output is growing but not yet complete.
Security and accuracy are top concerns: Seventy-three percent cite code accuracy as their primary concern, followed by security vulnerabilities (61%) and context understanding limitations (58%).
The learning curve is real: Developers who invest time in learning prompt engineering and tool-specific features report significantly higher satisfaction and productivity gains.
Our survey reveals a landscape defined by widespread adoption but uneven mastery. While AI coding tools are now commonplace, most developers have not yet embedded them deeply enough into their workflows to realize their full potential. The following exhibits illustrate key patterns in adoption, proficiency, use cases, and challenges.
Exhibit 1
AI coding tool adoption has reached near-universal levels among developers.
Frequency of AI coding tool usage among 200 surveyed developers
Exhibit 2
Self-reported proficiency levels reveal a significant skills gap.
How developers rate their own proficiency with AI coding tools
42% of developers consider themselves intermediate or advanced users— indicating significant room for skills development across the industry.
Exhibit 3
Code completion dominates, but developers are finding value across the workflow.
Primary use cases for AI coding tools, percentage of respondents
Exhibit 4
Accuracy and security concerns persist as primary adoption barriers.
Top challenges developers face when using AI coding tools
About the research
The online survey was in the field from October to December 2024, and garnered responses from 200 professional software developers across 12 countries representing a range of experience levels, company sizes, and technical backgrounds.
Participant demographics
- Experience level: 35% junior (0-3 years), 42% mid-level (3-7 years), 23% senior (7+ years)
- Company size: 28% startup (<50 employees), 31% mid-size (50-500), 41% enterprise (500+)
- Primary languages: JavaScript/TypeScript (68%), Python (54%), Java (32%), Go (18%), others (46%)
- Tools used: GitHub Copilot (72%), ChatGPT (68%), Cursor (34%), Claude (28%), others (41%)