Survey: Developers Are Using AI More… But Trusting It Less

84% of software developers report using or planing to use AI tools in their workflow, according to a big new Stack Overflow survey. That’s an increase from the 76% who said the same thing one year ago.

However, just one in three (33%) developers say that they trust the accuracy of AI tools — which is down from 43% in 2024.

In other words, AI has never been more widely used, yet an increasing number of developers are finding that they trust it less as their experience with it grows.

Fewer Developers Have Trust in AI

Stack Overflow’s 2025 developer survey compiled responses from a massive pool of over 49,000 developers in 177 countries. It’s one of the biggest annual surveys in the sector, making it a great yardstick for general AI acceptance.

Not only did the new survey find that developers’ trust in AI has dropped, but it also charted a decline in AI “favorability,” which now stands at 60%, down from 72% in 2024 and 77% in 2023.

The results are surprising, given how well generative artificial intelligence has saturated the business world by mid-2025.

“One of the most surprising findings was a significant shift in developer preferences for AI compared to previous years, while most developers use AI, they like it less and trust it less this year. This response is surprising because with all of the investment in and focus on AI in tech news, I would expect that the trust would grow as the technology gets better.” -Erin Yepis, Senior Analyst for Market Research and Insights at Stack Overflow, to VentureBeat

Has the Tide Turned Against AI?

Granted, the trust levels in 2024 were only 43% (barely up from the 42% that AI rated in 2023), so there was never a particularly high trust in AI from the start.

Still, a drop to 33% indicates that developer trust in AI is headed the wrong direction for the many big-tech players that have placed huge bets on the still-nascent technology in recent years.

Anecdotally, I can confirm that one editor I know has told me that they don’t think AI tools can accurately replicate a good writer’s output, but that they think AI could be useful for writing code.

Meanwhile, I’ve seen coders complaining about the ways in which AI inserts time-consuming bugs in the code it writes — even if they trust it to write the written word. In both cases, those with experience in their field say that AI can’t quite cut it.

Fixing AI Mistakes Takes Up Time

There’s no denying that AI tools can do simple tasks. But the complexities of software development might just be beyond the current technology, if this new survey is any indication.

Take the most common problem reported in the survey, for example. 66% of developers said they were dealing with “AI solutions that are almost right, but not quite.” The second place concern was raised by the 45% who said debugging AI code took more time than they wanted.

The exact question was “When using AI tools, which of the following problems or frustrations have you encountered?” and the third-most common response is an interesting one. 23.5% of respondents simply said: “I don’t use AI tools regularly.”

In the future, perhaps opting out of AI tools entirely will be the new hyped-up trend.