Go developers often use AI tools – but are not very enthusiastic

Contents

The team behind Google's programming language Go surveyed the community, and over 5000 Go developers responded: They frequently use AI tools for information retrieval and coding, but are only moderately satisfied with them. In contrast, Go developers are almost universally satisfied with the programming language itself: 91 percent indicated this.

The free-text responses provide insights into the reasons for its popularity. The Go team cites, among other things, that other languages are too complex, while Go is comparatively small and simple. Another response emphasizes the good scalability of the language, which is suitable for both individual programmers and large teams. Further responses highlight the tooling, the standard library, and the Go ecosystem positively.

The biggest hurdles for Go developers

The difficulties in Go development were also surveyed. The top 3 most frustrating aspects are “Ensuring Go code follows best practices and Go idioms”, “A feature I appreciate in another programming language is not present in Go”, and “Finding trustworthy Go modules and packages”. Many developers also find the documentation and testing of Go code, as well as learning Go modules, difficult.

AI tools with obstacles

Over half of the respondents use AI tools daily, while 29 percent either never use them or used them only a few times within the month before the survey. The most frequently used AI assistants are ChatGPT (45 percent), GitHub Copilot (31 percent), Claude Code (25 percent), and Gemini (20 percent). However, directly in fifth place with 17 percent is: “I do not use AI assistants or agents when writing Go code”.

The use cases for AI are diverse and go beyond writing code. For example, developers use it to learn new concepts, search for information, use AI for testing, and review their own written code. However, one-third do not want to review other people's code with AI assistance, and the respondents largely reject AI for deploying software to production.

Satisfaction with AI tools is mixed: only 13 percent are very satisfied, and 42 percent are somewhat satisfied. 16 percent are dissatisfied, and seven percent are very dissatisfied. The remaining 22 percent are neutral. The majority cite non-functional code as the main problem regarding AI development tools, while nearly a third also criticize the quality of functional AI-generated code.

Cross-language insights

This year, the Go team also asked what – besides Go – is the programmers' other favorite language. Python, Rust, TypeScript, C, and Java are at the top. The Go team sees potential here to create specialized tutorials, for example, “Error Handling in Go for Java Developers”. Part of such guides could also be integrated into code analyzers and thus made directly available within an IDE.

Data basis: Professional and experienced Go developers

The 2025 Go Developer Survey took place from September 9 to 30, 2025, and yielded 5379 valid responses. Of these, 87 percent work as developers. The respondents primarily use Go professionally (82 percent) and privately or for open-source projects (72 percent), and one-third already have more than six years of experience with Go.

Further information on the results can be found on the Go blog. The raw data is also expected to be released in the first quarter of this year. However, it will only contain the responses of individuals who have consented to this publication (82 percent).

(mai)