Programmers are using AI more than ever, but they do not like or trust the equipment very much, according to it 2025 stack overflow development survey,
A survey of around 50,000 developers found that 84% now use or plan to use AI tools in their workflows last year (76%). More than half of professional developers (51%) use these devices daily.
Also: Best AI for coding in 2025 (and what not to use)
Such figures may suggest that the programmer should love AI. However, only 60% expressed positive feelings towards AI devices, a ratio of more than 70% in both 2023 and 2024.
Destust is a defined subject of survey. In 2024, 43% of developers felt good about AI accuracy, and only 31% suspected. By 2025, 33% of the developers rely on the AI Tool Output, 46% expressed active mistrust, and only 3% said they highly trusted the results. In experienced professionals, the “excessive trust” figure only fell to 2.6%, with 20% strong doubt reporting.
Also: poor vibes: How an AI agent makes his way for disaster
In short, developers definitely use AI, but relying on technology to get a job is another case. And then the experience with a vibe programming project of Leader Jason Lekkin which disintegrated, took it with a production database, which can blame him?
Stack overflow studies also found that 66% of the largest single frustration cited by 66% developers, “AI is working with solutions that are almost right, but not enough,” who often leads to the second largest disappointment, at 45%, at 45%, “Debugging AI-Janit Code is more time-in-time.”
Bill Harding, K. CEO density And GuitclierNoted in gitclear AI Copilot Code Quality The study, which analyzed 211 million lines of the code, “developers rely on the current generation of AI assistants, as we rely on the previous generation, that is, not more.”
It is not just a programmer who does not rely on AI. A recent survey of over 1,100 Americans showed that only 8.5% said they “always trust” the information obtained from Google’s AI overview, and 21% said they had zero confidence in the ability to facilitate exact information for the surface. A new KPMG study found that, worldwide, 66% of people use AI, but only 46% Trust AI System,
Also: How to get rid of AI overview in Google Search: 4 Easy Ways
Junior developers who trust AI can make a big mistake. A popular blog post by an independent developer Namaiya Goel warned: “We are Trading deep understandingAnd when it looks great at the moment, we are going to pay for it later. ,
Concerns over AI-built technical loans are increasing. Harding warned that if companies continue to measure developer productivity by simple-brain metrics of committees written or the number of lines written, the AI-manual will be out of technical debt control. ,Leaders need to identify that more code is often worse,
Everyone said, Stack overflow also revealed in its survey that Openi’s GPT Models are the most popular big language models, with 82% of developers using AI which shows that they used them for development work in the previous year. Anthopropic Cloud sonnet The model came in second place, then Google’s Gemini flash,
Despite the rise of AI, when it comes to this integrated growth environment (IDE), the programmers still prefer visual studios (75%) and visual studio code (29%) on AI-first programming IDE. Old school, simple code editor, vim and notepad ++ are popular, even among the programmers who use AI. He said, Microsoft has proved to be a smart move to include Copilot in his equipment.
Too: Coding with AI? My top 5 tips to vetting its output – and stay out of trouble
As before, JavaScript, HTML/CSS, and Python maintain their position as the most widely used programming languages. Python, and perhaps due to the popularity of python-based generic AI library tensorflow and pitoch, is sought after developers, especially after adopting a new language. However, the rust, with an 83% approval mark, remains the most acclaimed language.
Looking forward, while AI is being adopted quickly, there are developers, if anything, the agents are being alert about handing over important tasks. A vast majority (75%) stated that human advice is still irreparable in the scenarios where they do not rely on the production of AI.
For AI agents, they have yet to reach mainstream acceptance. More than half of the survey respondents used simple AI tools, and 38% have no plans to adopt agents soon.
