42% of Developers Say AI Writes Half Their Code

Estimated reading time: 5 minutes · Last updated: 16 September 2026

BairesDev’s Q3 2026 Dev Barometer finds that 42% of respondents say AI now writes at least half of their code, up from 12% in Q3 2025. The survey polled 705 developers across more than 60 countries and collected answers from 41 enterprise CTOs. BairesDev’s figures show AI reduces the time spent producing code by roughly 13 hours per week on average, but those hours are being redirected into activities such as code review, debugging and learning new tools rather than lowering overall workload; BairesDev CEO Darren Shimkus called the shift “a layer above the code,” saying human validation has become the bottleneck, as first reported by VentureBeat.

Developers nearly doubled the time AI saves them in coding, to 13 hours a week. Not one of those hours came back.

Darren Shimkus, BairesDev CEO

Survey snapshot: sample, scope and the headline increase

BairesDev’s Q3 2026 Dev Barometer polled 705 developers across more than 60 countries and 41 enterprise CTOs. The developer sample is composed largely of applicants participating in BairesDev’s screening process rather than only the company’s employees, and the CTO sample is substantially smaller than the developer sample. The headline metric is a rise from 12% in Q3 2025 to 42% in Q3 2026 in the share of developers who say AI writes at least half their code. That is a change of 30 percentage points between the two quarters.

The report sits alongside several large-scale surveys that show the same pattern: JetBrains’ 2026 Developer Ecosystem Survey sampled more than 15,000 professional developers and found 90% use AI coding agents at least weekly, with 68% using them daily; Stack Overflow’s 2025 Developer Survey drew more than 49,000 responses and reported 80% using AI tools but only 29% trusting their accuracy. Those external datasets give context to BairesDev’s sample, which mixes applicants and practitioners.

Where developer time actually went: coding saved, review and learning rose

BairesDev reports that AI now saves developers 13 hours of coding per week, nearly double the roughly seven hours reported a year earlier. The extra time has not translated into idle capacity: only 21% of respondents say they now spend more than half their workweek writing new code from scratch, while 67% report spending more time reviewing AI-generated code and 52% report spending more time debugging problems introduced by AI.

The survey also finds a jump in formal upskilling: developers report averaging nine hours per week learning AI tools and other new technologies, up from four hours a year ago. That five-hour increase in weekly learning is comparable in scale to the additional review and debugging work respondents describe, and BairesDev’s CEO Darren Shimkus summarised the shift by saying, “Developers nearly doubled the time AI saves them in coding, to 13 hours a week. Not one of those hours came back.”

Enterprise reaction: more validation, not full delegation

The CTO responses indicate organisations are funding human checkpoints rather than delegating production decisions to agents. Among the 41 CTOs surveyed, 78% said they had increased spending on code review, quality assurance and validation to support AI-generated work. At the same time, only 7% of developers said the decision to ship code had been entirely delegated to AI without their input.

Shimkus told VentureBeat that enterprise customers are treating developer explainability as non‑negotiable for production systems: if an engineer cannot explain what AI-generated code does and why, the organisation will not permit that engineer to check it in. That distinction—greater autonomy for prototypes, strict human review for production—helps explain why spending has shifted into review processes even as AI generates larger code volumes.

How this reshapes skills, pay and security priorities

The barometer shows AI fluency is already affecting compensation and promotion. Among developers who received raises in 2026, 29% identified AI tool fluency as the driver, followed by system design and architecture at 20% and human skills such as communication and mentorship at 15%. Yet only one in four CTOs said they were actively investing in human‑skills development, a mismatch BairesDev flags as a strategic gap.

Security and governance are rising on the agenda as agentic systems take on actions beyond code generation. BairesDev points to external research reported by ITPro that finds limited use of deployment gates and governance around agents. The company’s advice—go deep in security and in one AI tool rather than many—echoes Shimkus’s warning that organisations can buy tools but not necessarily the judgment to apply them safely at scale.

How the trend could evolve

The bottom line

BairesDev’s Q3 2026 Dev Barometer shows a rapid increase in AI-generated code volume, but the practical effect is a reallocation of developer effort rather than freed capacity. Engineers report an extra 13 hours of AI-saved coding time per week, yet those hours are absorbed by review, debugging and learning; organisations are responding by increasing spending on validation. For employers the implication is clear: tools alone do not replace judgment—investment in review pipelines, security controls and human skills will determine whether AI raises team productivity or simply shifts costs.

What to watch

• Watch for BairesDev’s next Dev Barometer release; no date has been set.
• Watch for enterprise guidance or public policies on deployment gates for agentic development; no date has been set.
• Watch for major vendors or large enterprises publishing standards for AI-generated code validation; no date has been set.