Stack Overflow's Q&A Traffic Collapses, But AI Data Licensing Keeps Company Alive

Stack Overflow’s public forum has effectively cratered under pressure from ChatGPT, Copilot, Cursor, Claude, and Gemini. Question volume last month dropped to roughly 6,866 — about where it sat at launch in 2008 — as developers route routine coding questions to LLMs instead of the site.

The company itself, however, is doing better financially than the engagement metrics suggest. Annual revenue has roughly doubled to $115 million while losses shrank from $84 million in FY2023 to $22 million, helped along by mass layoffs and a pivot away from ad-supported traffic. The new business is enterprise: Stack Internal, a generative-AI product trained on the site’s archive, now serves 25,000 companies, and Stack Overflow licenses its corpus to AI labs in a model similar to Reddit’s nine-figure data deals.

The arrangement is circular and fragile. LLMs need human-curated coding knowledge to stay useful, and Stack Overflow holds one of the largest such corpora — but that corpus is aging as the community that produced it shrinks, with complex questions still landing on the site only because nowhere else handles them. CEO Prashanth Chandrasekar is betting the long tail of hard problems plus historical depth is enough to sustain the licensing business, even as the forum that generated the data continues to wither.