AI coding statistics 2026, by the numbers.

In 2026, 90% of professional developers use AI coding agents at work at least weekly and 68% use them every day, according to JetBrains’ survey of more than 15,000 developers from May to July 2026. The same developers say agents fully write about 47% of their code, yet trust stays low: in Stack Overflow’s latest published survey, 46% of developers distrust the accuracy of AI tools and 33% trust it.

Below are the most useful AI coding statistics of 2026, each linked to its primary source with its date and sample. We checked every source on September 28, 2026. Most large developer surveys come from companies that sell developer tools, so we name who ran each one.

01How many developers use AI coding tools in 2026?

Almost all of them. Every large survey since mid-2025 puts regular AI use by developers at 84% to 90%, and daily use of coding agents is now the norm among professionals. Checked on September 28, 2026; each row links its source:

The trend is steep. Stack Overflow’s 2025 survey found only 14.1% of developers using AI agents at work every day, and 37.9% with no plans to use them at all. A year later, JetBrains measured daily agent use at 68%. The two surveys ask different people different questions, so read that as a direction, not a precise jump.

Stack Overflow had not published its 2026 results when we checked: its 2026 results page returned “not found” on September 28, 2026.

02Which AI coding tools do developers use most?

Claude Code is the most used AI coding tool at work in 2026, used by 39% of professional developers in JetBrains’ May to July survey, and by 47% in the United States. GitHub Copilot, the leader a year earlier, fell to 21%.

Share of professional developers using each tool at work. January figures come from JetBrains’ AI Pulse survey of 10,000+ developers; May to July figures from its Developer Ecosystem Survey 2026 of 15,000+. JetBrains makes its own AI tools, which appear in the table. Checked on September 28, 2026.

Codex grew fastest, about fivefold between January and mid-2026, and awareness of it rose from 27% to 65% of developers. For what each tool does and costs, see our comparison of the best AI coding agents in 2026.

03What percentage of code is written by AI?

Developers say AI agents fully write about 47% of their code, according to JetBrains’ survey of 15,000+ professional developers from May to July 2026. Another 38% is written with some AI help. These are self-reported estimates of the code each person produced the previous month, not measured counts.

• About 22% of developers get more than 80% of their code from agents, and about one in five writes no code at all without AI help.

• Go, JavaScript and TypeScript developers report the highest agent share, 54% to 55% on average. C and C++ developers write the most by hand.

• Senior developers lead: about a quarter of them get more than 80% of their code from agents, a larger share than juniors.

Sonar, a code quality vendor, asked 1,149 developers in October 2025 and got a similar answer: 42% of the code they commit is AI-generated or significantly AI-assisted. They put the 2023 share at 6% and expect 65% by 2027.

The best-known company figure in writing is older. In October 2024, Google CEO Sundar Pichai said more than a quarter of all new code at Google was generated by AI, then reviewed and accepted by engineers.

04Do developers trust AI-generated code?

Mostly not. In Stack Overflow’s 2025 survey, 46% of developers said they distrust the accuracy of AI tools and 33% trust it, while only 3.1% trust it highly. Experienced developers are the most wary: 2.6% highly trust AI output and 20% highly distrust it.

The gap between 96% doubting and 48% always checking is where bugs slip through. Our guide on how to review AI-generated code gives a method that does not require reading every line.

05Does AI actually make developers faster?

The best evidence says: probably yes now, but by less than developers believe. In METR’s randomized trial of 16 experienced open-source developers in early 2025, tasks took 19% longer when AI was allowed, even though the developers expected a 24% speedup and afterward believed they had gained 20%.

METR ran a larger follow-up from August 2025 and reported in February 2026 that tasks now took an estimated 18% less time with AI for returning developers and 4% less for new ones. Both ranges include no effect, and METR calls the data weak evidence. Between 30% and 50% of developers said they held back tasks they did not want to do without AI, which likely hides the biggest gains.

Self-reports run much higher. In a METR survey of 349 technical workers from February to April 2026, the median person said AI made them 3 times faster and their work 1.4 to 2 times more valuable. METR warns that people overestimated their speedup by 40 percentage points in its 2025 trial.

Whether faster coding changes hiring is a separate question. Our look at what the 2026 data says about AI and jobs covers the labor statistics.

06How often does AI-generated code contain security flaws?

In Veracode’s latest tests, 45% of AI code generation tasks produced code with a known security flaw, a rate that has barely moved in two years. Veracode, which sells security testing, ran 80 coding tasks in Java, JavaScript, C# and Python and has tested more than 150 models to date. Its March 2026 update found:

Source: Veracode Spring 2026 GenAI Code Security Update, March 24, 2026, checked on September 28, 2026.

Developers see the same problem from the inside. In Sonar’s survey, 88% reported at least one negative effect of AI on their code, most often code that looks correct but is not reliable (53%). Our checklist for securing AI-generated code lists the flaws to scan for first.

07FAQ

Read next: should you still learn to code in 2026, our AI agent statistics for company adoption beyond coding, or the coding agent leaderboard for benchmark scores.