The results from StackOverflow’s 2025 Developer Survey paint an interesting picture of software engineers’ trust in AI tools:
trust in the accuracy of AI has fallen from 40% in previous years to just 29% this year. We’ve also seen positive favorability in AI decrease from 72% to 60% year over year.
Despite this, adoption has increased to a high level:
AI tool adoption continues to climb, with 80% of developers now using them in their workflows.
We’re at a point where managers are increasingly expecting developers to leverage AI in their day-to-day work. In my opinion, these tools are highly valuable, and I use them regularly.
LLMs have become quite good at “one-shotting” a new feature or even an entire app. But when it comes to refining, debugging, and dialing in the last 20%—this is where they struggle. It actually mirrors the stages engineers tend to go through. This last portion takes the most thinking, troubleshooting and the most time. It’s where AI becomes the most inaccurate.
The initial broad strokes of building UI, laying out the architecture and basic functionality tend to be the easiest to deal with. AI is like a rocket booster for this. The later phases may be a different story.