The Double-edged Effect of Banning Generative AI on Online Question-and-Answer Communities: Evidence from Stack Exchange
2026-07-06 • Computers and Society
Computers and Society
AI summaryⓘ
The authors studied what happens when online Q&A communities ban AI-generated content, like answers from ChatGPT. They found that banning AI leads more people to ask questions, but fewer questions get answered well and quickly, especially in non-STEM communities. The authors suggest this happens because AI is less reliable for some topics and social interaction needs differ across communities. After the ban, people tend to write more detailed and socially engaging questions and answers. These findings help people who manage or moderate online Q&A sites understand the effects of such bans.
generative artificial intelligenceAI-generated contentStack Exchangedifference-in-differencesknowledge seekingcontribution efficiencynon-STEM communitieslarge language modelsonline communitiessocial interactivity
Authors
Yuanhong Ma, Qinglai He, Xitong Li, Lynn Wu
Abstract
We investigate how banning generative artificial intelligence-generated content (AIGC) affects knowledge seeking, knowledge contribution, and contribution efficiency in online question-and-answer communities. After the launch of ChatGPT in late November 2022, several Stack Exchange communities implemented official bans on AIGC over concerns such as less reliable and socially engaged content. Leveraging data from the full network of Stack Exchange communities, we employ a difference-in-differences (DID) approach to examine the impacts of these bans. Our results reveal a double-edged impact: while the AIGC ban increases knowledge seeking, as evidenced by a higher volume of posted questions, it simultaneously reduces contribution efficiency, reflected in a lower proportion of questions receiving satisfactory answers within the expected time frame. Notably, these impacts are only evident in non-STEM communities. We take a socio-technical perspective to explore information reliability and social interactivity as two plausible underlying factors driving the observed changes. Our mechanism exploration reveals that the AIGC ban spurs question volume in topics where AIGC is less reliable and where social interaction is highly expected. In contrast, the ban hampers answer efficiency in communities where LLMs are capable of producing reliable answers and where social interactivity is minimal. Additionally, our results indicate the increased human involvement from knowledge seekers and contributors following the ban. They adapt their behavior by posting questions and answers that are more informationally rich and socially engaging. Overall, our findings offer actionable implications for platform managers, community moderators, and policymakers of online Q&A communities.