- Generative AI may have inadvertently driven away many top‑tier expert contributors from platforms like Stack Overflow, as users turn to tools trained on that very expertise.
- The problem arises when contributors sense their effort goes unrewarded, since AI can deliver comparable answers much more quickly.
- This trend extends beyond coding forums, threatening to affect classrooms, corporate workplaces, and scientific communities alike.
A study from the University of Auckland tracking Stack Overflow’s decline highlights a troubling pattern: the platform’s most skilled contributors are exiting en masse.
While AI helps narrow the divide between novice and mid‑level programmers and elite talent, it may also hasten the departure of those experts, who feel their contributions are less valued.
Since ChatGPT launched in 2022, Stack Overflow’s monthly question volume has dropped by almost 76%, signaling that newcomers and longtime users alike are leaving the site.
Latest Videos From
A Wider Issue Beyond Stack Overflow
While Stack Overflow’s decline had many causes, many users believe the site and its top contributors exhibited a degree of arrogance.
Combined with moderation many described as self‑righteous, this drove users toward alternatives, leading them to abandon the platform.
ChatGPT and similar AI tools grew more adaptable, eventually serving as search engines for programmers tackling routine queries, while also improving at handling syntax‑related questions.
This, in turn, lowered the volume of questions on the site; even a later ban on generative AI could not stem the loss of answerers, a gap that may be hard to fill long‑term.
Researchers warn the problem is no longer limited to coding forums; it could spread to classrooms, offices, and other research settings, where distinguishing low‑effort AI replies from expert answers is increasingly difficult.
“If anyone can produce a decent answer with AI, some may wonder why they should bother sharing their expertise,” said Dr. Kenny Ching, publisher of the study.
He described this as ‘signal compression,’ where distinguishing expert from novice outputs becomes difficult, and contributing as a subject‑matter expert feels less rewarding when AI can address the same topics.
A simpler question remains: if AI learns from user‑generated content and that pool is shrinking on sites like Stack Overflow, where will the impending knowledge reset lead AI capabilities?
Future AI models won’t necessarily become less capable, but they may seek alternative training data—such as Slack threads, Discord conversations, or users who still pose the same coding questions they once asked on Stack Overflow.
Whether this shift replaces departing experts or simply makes AI more error‑prone due to its feedback loop remains an open question, especially as society struggles to tell AI‑generated from human‑produced answers.
