ChatGPT cut Stack Overflow's question volume by about 25% in six months: the peer-reviewed measure of an AI-disrupted knowledge platform, and the 28% layoff that followed

Verification status: PENDING — the checker verified the headline figure verbatim against the peer-reviewed PNAS Nexus study and the layoff against independent press (VentureBeat, SiliconANGLE); confidence 1.000, awaiting the owner’s sign-off before the green badge.

The problem

Stack Overflow is the public question-and-answer archive that a generation of programmers learned from: a developer hits an error, searches, and finds a human-answered question. Its problem was not a product it shipped but a free tool someone else did. When a general-purpose chatbot can answer the same coding question privately and instantly, the incentive to post that question in public, and for someone else to answer it, weakens. The question this case file settles is whether that displacement is real and measurable rather than merely asserted, and a peer-reviewed study answered it directly by comparing Stack Overflow’s activity against platforms less exposed to ChatGPT (source). The study frames its own stakes plainly: it measures posting activity on Stack Overflow “against that of four comparable Q&A platforms” chosen because “their users experience difficulties” reaching ChatGPT or because the subject matter is one where ChatGPT is weaker (source).

What was built

There is no Stack Overflow-built AI system at the centre of this case; the system doing the displacing was built by someone else. It is OpenAI’s ChatGPT, released, as the paper dates it, on “2022 November 30,” after which “posting activity decreased sharply” (source). What was “built” here, in the sense this registry cares about, is the measurement: a difference-in-differences design that isolates ChatGPT’s effect by using Russian and Chinese Q&A platforms and mathematics forums as counterfactuals, precisely because they were less able or less likely to use the tool. That design is what turns a widely-assumed disruption into a quantified one, and it is why the finding is load-bearing rather than anecdotal (source).

The outcome

The measured effect was large and fast. The peer-reviewed paper reports that “Within 6 months of ChatGPT’s release, activity on Stack Overflow decreased by 25% relative to its Russian and Chinese counterparts, where access to ChatGPT is limited, and to similar forums for mathematics, where ChatGPT is less capable” (source). In raw terms the authors describe the same fall on the platform itself: after ChatGPT’s release “posting activity decreased sharply, with the weekly average falling from around 60,000 posts to 40,000 within 6 months” (source). The authors are explicit that they treat this as a floor, not a ceiling: they “estimate a 25% decline to posts on Stack Overflow relative to the counterfactual platforms within 6 months of ChatGPT’s release,” adding that they “interpret this as a lower bound” (source).

Two secondary findings sharpen what kind of loss this was. It was a loss of quantity, not quality: the study reports “no significant change in post quality, measured by peer feedback, and observe[s] similar decreases in content creation by more and less experienced users alike” (source). And it tracked exposure to the tool: “The decline is larger for posts related to the most widely used programming languages,” the languages where ChatGPT is most capable (source).

The business consequence followed within the same window, though it must be read carefully. In October 2023 Stack Overflow “is laying off 28% of its workforce, CEO Prashanth Chandrasekar announced today” (source), a cut corroborated by SiliconANGLE’s contemporaneous report that the company let go 28% of its workforce in a “cost reduction push” (source). Crucially, Stack Overflow itself did not blame AI: Chandrasekar said “this year we took many steps to spend less” but that “Unfortunately, those changes were not enough,” and cited “ongoing threats to customer budgets shifting due to the macroeconomic pressures impacting the entire tech industry” (source). The layoff is documented fact; the causal, quantified link between ChatGPT and reduced Stack Overflow activity is what the peer-reviewed study supplies, and the two should not be conflated.

Weakest load-bearing source. The headline claim is unusually strong for this registry: it rests on a peer-reviewed study in PNAS Nexus, an independent academic source, not on a vendor or on the platform’s own marketing (source). The genuinely weaker point is the causal framing of the layoff. The 28% cut is well-sourced, but Stack Overflow attributed it to profitability and macroeconomic pressure, not to AI, so this page treats the layoff as corroborating context for a disrupted business rather than as itself proof of AI causation (source). The 25% activity decline, by contrast, is the number the independent study was designed to measure. No confirmation was sought from Stack Overflow or the authors; every source is already public.

How this was verified. Method: every load-bearing figure was retrieved live this session, grepped for the exact string, and saved to sources/. The ~25% activity decline, the ~60,000-to-~40,000 weekly-post fall, the “no significant change in post quality” finding, the “lower bound” characterisation and the larger effect for widely-used programming languages are quoted verbatim from the peer-reviewed paper (R. Maria del Rio-Chanona, Nadzeya Laurentsyeva, Johannes Wachs, “Large language models reduce public knowledge sharing on online Q&A platforms,” PNAS Nexus 3(9):pgae400, published 11 September 2024). Its full text was read through PubMed Central (saved to sources/pnasnexus-pgae400-pmc.html, Wayback 20260902235917) and cross-checked against the Oxford Academic canonical page (Wayback 20260712184706). The October 2023 layoff of 28% of the workforce and CEO Prashanth Chandrasekar’s verbatim stated reasons are quoted from VentureBeat (saved to sources/venturebeat-so-layoffs-28pct.html, Wayback 20250729172518) and corroborated by SiliconANGLE (saved to sources/siliconangle-so-layoffs-28pct.html, Wayback 20260903000148). No confirmation was sought from Stack Overflow or the study’s authors. Checked 2026-09-03 (checking round 1).

Related case files

• ChatGPT wiped nearly $1 billion off Chegg in a day, and by 2025 the edtech had cut ~45% of its workforce — the closest analogue: another incumbent whose business rested on answering questions students now put to a chatbot, told from its own disclosures.
• A controlled experiment found GitHub Copilot let developers finish a task 55.8% faster — the flip side of the same shift, quantifying the productivity of the AI coding tools that reduced the need to ask Stack Overflow in the first place.
• CNET quietly used an AI tool to write articles, then corrected 41 of 77 and paused it — another business built on human-produced answers colliding with generative AI, where the damage was to trust rather than volume.

Sources

1. R. Maria del Rio-Chanona, Nadzeya Laurentsyeva, Johannes Wachs (PubMed Central) · Large language models reduce public knowledge sharing on online Q&A platforms · PNAS Nexus 3(9):pgae400, 11 September 2024 · https://pmc.ncbi.nlm.nih.gov/articles/PMC11421660/ — Tier 1 (primary, peer-reviewed independent academic study; the ~25% activity decline “within 6 months of ChatGPT’s release,” the ~60,000-to-~40,000 weekly-post fall, “no significant change in post quality,” the “lower bound” characterisation; full text saved to sources/pnasnexus-pgae400-pmc.html, Wayback 20260902235917).
2. R. Maria del Rio-Chanona, Nadzeya Laurentsyeva, Johannes Wachs (Oxford Academic) · Large language models reduce public knowledge sharing on online Q&A platforms · PNAS Nexus 3(9):pgae400, 11 September 2024 · https://academic.oup.com/pnasnexus/article/3/9/pgae400/7754871 — Tier 1 (primary, peer-reviewed; canonical journal page cross-checking the PMC full text, including “The decline is larger for posts related to the most widely used programming languages”; Wayback 20260712184706).
3. VentureBeat · Stack Overflow confirms layoffs affecting 28% of workforce · 16 October 2023 · https://venturebeat.com/programming-development/stack-overflow-confirms-layoffs-affecting-28-of-workforce/ — Tier 2 (independent technology journalism; the 28% layoff and CEO Chandrasekar’s verbatim reasons, “took many steps to spend less,” “not enough,” and the “macroeconomic pressures impacting the entire tech industry” attribution; saved to sources/venturebeat-so-layoffs-28pct.html, Wayback 20250729172518).
4. SiliconANGLE · Stack Overflow lets go 28% of its workforce in cost reduction push · 16 October 2023 · https://siliconangle.com/2023/10/16/stack-overflow-lets-go-28-workforce-cost-reduction-push/ — Tier 2 (independent technology journalism; second independent source corroborating the 28% October 2023 layoff; saved to sources/siliconangle-so-layoffs-28pct.html, Wayback 20260903000148).