AI Coding Tools Feel Addictive to 80% of Developers: Survey

What happened

A new Coddy Developer Survey has found that four in five developers say their use of AI coding assistants feels more compulsive than genuinely productive, with respondents describing the experience in terms usually reserved for addictive behaviour rather than workplace tooling. According to reporting on the survey by ZDNET, developers frequently reached for phrases such as "I can't stop" when describing their reliance on AI code-generation tools.

The survey frames this as an emerging, distinct form of burnout — not the traditional fatigue of long hours or high workload, but one driven by the constant, low-effort dopamine loop of prompting an AI assistant, accepting or rejecting suggestions, and repeating the cycle. Rather than easing cognitive load, the tools appear in many cases to be creating a new kind of dependency that developers themselves recognise as unhealthy.

Why it matters

This is a significant data point in the wider debate about AI adoption inside technical teams. Much of the enterprise conversation about AI coding assistants has focused on productivity gains — faster shipping, fewer bugs, lower onboarding time for junior engineers. This survey suggests the human-experience side of that equation is more complicated: the same features that make these tools frictionless and rewarding to use may also be encouraging compulsive, low-value engagement rather than deliberate, high-quality work.

For organisations rolling out AI coding tools at scale, this points to a genuine employee-experience and workforce-wellbeing question sitting alongside the productivity one. Engineering leaders measuring success purely through adoption or usage volume may be missing a signal that heavy, habitual use is not the same as effective use.

By the numbers

• 80% of developers surveyed said they find AI coding tools more addictive than genuinely helpful, per the Coddy Developer Survey.

The Renascence take

The instinct in most transformation programmes is to treat AI tool adoption as an unambiguous win the moment usage numbers climb. This survey is a useful corrective: usage and value are not the same thing, and behavioural design — the interaction patterns that make a tool feel rewarding to use — can just as easily produce compulsive habits as genuine efficiency.

What's being missed here is that AI coding assistants are, functionally, variable-reward interfaces: a prompt, a suggestion, a small hit of "did that work," repeated endlessly. That is the same behavioural mechanic behind most habit-forming consumer apps, now sitting inside professional workflows. A customer-obsessed — and employee-obsessed — operator shouldn't ask "how much are our developers using this?" but "are they producing better outcomes with less friction, or are they simply looping?" The fix isn't banning the tools; it's redesigning how success is measured and paced, building in deliberate checkpoints, and treating engineering wellbeing as a service-design problem, not just an HR one.