Skip to content
Stack Overflow Dev Survey 2026

AI

OverviewFull data

AI moves from on-the-side to built-in

Coding assistants and agents have pulled ahead as the dominant use case for AI (66%), outpacing even general-purpose LLM-powered chatbots (63%). It’s a sign that AI has become less sidekick and more infrastructure, built right into the tools technologists already use. The overwhelming majority of respondents now use coding assistants and agents daily (73%), with nearly a third of daily users spending upwards of four hours.

AI usage

31% of daily AI users use it for 4+ hours/day

Of the over 70% of respondents who use AI daily, 31% use it over four hours per day.
View and share
4+ hours/day 31%, 1-2 hours/day 25%, 2-4 hours/day 24%, <1 hour every day 12%, <1 hour and not every day 7%, None 1% · All Respondents · n = 10,92631%4+ hours/day
RespondentsPercent
4+ hours/day3,37830.9%
1-2 hours/day2,76525.3%
2-4 hours/day2,63624.1%
<1 hour every day1,33512.2%
<1 hour and not every day7637.0%
None490.5%

Despite plenty of societal concern about AI, survey respondents retain a fairly positive view of the technology.

When asked about overall attitude toward AI tools, favorable views outnumber unfavorable ones by almost 3 to 1. But “favorable” doesn't mean “unconditionally positive.” Developers aren't simply handing AI the keys to important decisions. When asked when it’s acceptable to trust AI, 48% indicate they trust AI when they can easily validate the answers and only 16% extend that trust to most tasks that are not important work decisions. The sweet spot where most AI users land is trusting AI when they understand what it’s working on, such as with familiar code (67% currently use AI with this) or when debugging or refactoring (61% currently use AI with this). That’s why human-in-the-loop is the dominant mode of working with these tools.

“This is where I personally feel the importance of human judgment. If you leave it up to chance, if you leave it up to the roll of the dice of the LLM, you're not gonna get the specificity. Everything I've been doing and everything my team's been doing is imagining AI as an exoskeleton, like a Tony Stark augmentation, as opposed to an empty Ultron with nobody inside. Which nobody wants.”

The right tool for the job

The same caution shows up in how people choose (and choose against) certain tools. Developers switching to new AI tools are doing it mostly because the output is better (25%) or because their organization required it (11%).

AI at work

25% switch AI tools due to better output quality

Seeking superior output quality is the single largest reason developers switch AI tools, though a third (34%) retained their original tool choices over the past year. For all the concerns over token costs, only 6% switched because of costs.
View and share
No 34%, Better output quality 25%, My organization required it 11%, Better integration 7%, Yes — lower cost 6%, Better agent support 4%, Another reason 4%, The old tool became limited 3%, and 4 more · All Respondents · n = 13,14425%Better output quality
RespondentsPercent
No4,42333.7%
Yes — better output quality3,30525.1%
Yes — my organization required or recommended a different tool1,39410.6%
Yes — better integration with my workflow or development environment9757.4%
Yes — lower cost8356.4%
Yes — better support for agents, automation, or coding workflows4873.7%
Yes — another reason (write in)4663.6%
Yes — the previous tool became too expensive, limited, or unavailable4333.3%
Yes — better security, privacy, or compliance2832.2%
Yes — access to a preferred model or provider2802.1%
Yes — for a specific feature1791.4%
Not applicable — I do not use AI tools for work840.6%

What are they using? Claude Code (66%) and GitHub Copilot (59%) are the leaders for coding agents. Respondents are using LangChain (30%) and OpenAI Agents SDK (24%) for orchestration and Make (39%) and n8n (33%) for automation. AI memory and agent orchestration tools show a notable gap between current adoption and stated interest, suggesting that the category is still finding its footing. Coding agents, meanwhile, are a more mature market with clear leaders emerging.

That selectivity becomes very clear once you look at where AI is actually being put to work. Basically, developers trust AI with machines but not with people. AI use is common, and commonly accepted, across code generation, debugging, testing, and understanding unfamiliar codebases. Even secondary review functions—analyzing telemetry, reviewing pull requests, hunting for security issues—are seeing meaningful AI adoption.

The human difference

But the moment the task shifts toward human communication or creative output—writing emails, generating multimedia, anything requiring a feel for what people want—usage falls off a cliff. The same split shows up in perceived usefulness: technical tasks like implementation and debugging are both useful or very useful to 70%+ of survey respondents, while communication and design lag behind, landing in “slop” territory with 35% and 49% respectively finding AI useful or very useful in those areas.

The top 5 tasks that respondents are currently using AI or AI agents for
Generating code in a familiar area 69%, Debugging, troubleshooting, or refactoring code 64%, Answering straightforward technical questions 59%, Writing or improving tests 58%, Generating code in an unfamiliar area 56% · All Respondents · n = 12,547Generating code in a famili…69%Debugging, troubleshooting,…64%Answering straightforward technical q…59%Writing or improving tests58%Generating code in an unfamiliar area56%
RespondentsPercent
Generating code in a familiar area8,40269.3%
Debugging, troubleshooting, or refactoring code7,68463.8%
Answering straightforward technical questions7,01759.4%
Writing or improving tests6,91558.1%
Generating code in an unfamiliar area6,70756.0%
The inevitable takeaway is that technologists see AI as very good at technical problems with checkable answers, and considerably less good at solving the other everyday problems of work. A significant chunk of respondents who avoid AI completely do so out of concern for their own skills, privacy, or the technology’s enormous ethical and environmental footprint.

AI at work

6 out of 10 users cite a reason to avoid using AI at work or school, losing job skills is a top concern.
View and share
I don't avoid AI 41%, Losing skills to AI 17%, Morals/ethics 15%, Privacy or security 13%, Environmental concerns 8%, Other 3%, Lack of training 3% · All Respondents · n = 13,85717%Losing skills to AI
RespondentsPercent
I do not avoid using AI at work or school5,72841.3%
Concern over losing job skills or training an AI to replace you2,36717.1%
Moral or ethical concerns2,10815.2%
Issues with privacy or security1,80213.0%
Environmental concerns1,1298.2%
Other, please specify:3692.7%
Difficulty or lack of training3542.6%
Explore AI data

What’s next?