Programmers are starting to trust AI more – but still don’t entirely believe it won’t come for their jobs
- 58% use AI to write code daily/often/sometimes, but 28% never use it at all
- Many lack trust in output and think AI-generated content could be incorrect
- GitHub Copilot and ChatGPT are the most popular tools, Claude rising
The latest annual global C++ developer survey has revealed many programmers are actually rapidly increasing their use of AI, however overall trust in the technology still remains low.
The data showed that writing code and tests are the most common use cases for artificial intelligence, while debugging, code review and identifying performance issues are less of a priority, implying that AI tools are being used as a first step, and human oversight remains crucial.
However, while many developers use these platforms across these use cases a couple of times a week, many still report never using AI at all.
Developers aren’t ready to fully trust AI
For example, around 58% of developers use AI to write code ‘almost every day’, ‘often’ or ‘sometimes’, 14% use it ‘rarely’, and 28% never use it.
While adoption, or at least experimentation, is growing, there are still some concerns at play. More than three in four (78%) worry that AI-generated content is incorrect – a lack of trust in output (70%), poor understanding of relevant context (51%) and data privacy concerns (50%) are also among the most popular setbacks. Many also object to AI on ethical and environmental grounds, particularly as tech companies continue to cut jobs that are seeing the biggest productivity boosts from AI.
Responding about where AI’s good, and where it’s not so good, the 1,400+ respondents generally felt that AI lacks in large, complex C++ projects, safety-critical code, architecture-level decisions and highly contextual debugging. However, it’s proving a hit across understanding unfamiliar or legacy code.
GitHub Copilot remains the most-used code-specific AI tool (53%), but Claude Code (44%) is gaining market share. OpenAI’s Codex (14%) is far less popular, however ChatGPT (53%) leaders the way in terms of general-purpose AI assistance, with Gemini (39%) and Claude Chat (30%) falling behind.
By usage patterns alone, it looks like developers are keen to use AI as a supportive tool and not as a replacement, but AI also has a long way to go to prove its worth in programming, where a simple error could create a major security issue.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.