Are Large Language Models Killing Stack Overflow? Data Shows the Decline

Four months ago we asked whether large language models would make Stack Overflow irrelevant; now the data confirms they have.

Since then the site’s question volume has been steadily falling. The latest figures show a sharp decline, with the current monthly question count now comparable to the level when Stack Overflow launched in 2009.

Now, let’s look closely at the data:

2014: Question count began to drop while moderator efficiency improved, leading to faster closures and removal of low‑quality posts.

Around that time I stopped asking questions because the site felt unwelcoming.

March 2020: Pandemic lockdowns caused a traffic surge; developers turned to search and Stack Overflow for help.

June 2020: Question volume started decreasing again, faster than before.

June 2021: Stack Overflow was sold to private‑equity firm Prosus for $1.8 billion.

November 2022: After ChatGPT launched, question volume dropped rapidly. ChatGPT answers faster, is trained on Stack Overflow data, and provides polite, comprehensive replies.

May 2025: Monthly question volume is as low as it was in 2009.

Even without large language models, Stack Overflow might have become irrelevant due to policy changes and other factors starting in 2014, but LLMs have undeniably accelerated its decline.

It’s sad to see a platform that helped countless developers lose relevance, yet I believe developers will continue to gather in new spaces—whether web servers, WeChat groups, or other formats—to help each other.

Author: 场长