Wikipedia is caught in a textbook extraction loop. Generative AI companies — Google, OpenAI, Microsoft, Meta, Amazon — train their models on Wikipedia's five billion volunteer-written words, then serve those answers directly to users who never click through to the source. The result is a slow strangulation: fewer readers means fewer editors means a degrading encyclopedia means worse training data. Everyone loses, but the volunteers lose first. Helen Choi, writing in AI and Ethics and summarized on the LSE Impact Blog, frames this as a structural injustice under Iris Marion Young's Social Connection Model. The argument is straightforward: academics participate in knowledge systems, Wikipedia is a load-bearing element of those systems, and participation creates obligation. It's not that academics broke Wikipedia — it's that their position in the knowledge infrastructure gives them unique capacity and therefore unique duty to defend it. The practical proposals are concrete and already partially in motion. Academics can edit Wikipedia directly, leveraging subject-matter expertise to close content gaps. They can cite Wikipedia in scholarly work, signaling its legitimacy. And they can assign Wikipedia editing as coursework — the Wiki Education Foundation reports that over 152,000 students at 8,000 institutions have contributed 125 million words since 2010, with about 19% of Wikipedia's new active editors emerging from these programs. The student pipeline is the most structurally interesting piece. Students report higher motivation writing for a real audience, and instructors get an AI-proof assignment since Wikipedia bans machine-generated content. It's a rare intervention that simultaneously serves pedagogy, public knowledge, and platform sustainability. The problem is that institutional incentive structures don't reward it — public service contributions to Wikipedia go largely unrecognized in tenure and promotion evaluations. The extraction dynamics here are unusually clear. AI companies capture the commercial value of volunteer labor through either unauthorized scraping or paid API access — Microsoft, Amazon, and Meta pay Wikipedia for high-speed data feeds. The volunteers who created that value receive nothing. The readers who previously sustained the volunteer community are being redirected. And the companies doing the redirecting are the same ones profiting from the content. It's a circle, and it only goes one direction. Choi's argument has a structural limitation she acknowledges implicitly: academics alone cannot fix a problem created by companies with trillion-dollar market caps. The institutional reforms she calls for — recognizing Wikipedia contributions in academic evaluations — face the same inertia that has resisted broadening tenure criteria for decades. The 152,000-student pipeline is impressive but represents a fraction of global higher education enrollment. What makes this piece matter is not its novelty — the AI-is-killing-Wikipedia argument has been circulating for two years — but its attempt to assign specific responsibility to a specific group with specific capacity. Most commentary on AI extraction treats it as a weather event. Choi treats it as a governance failure with identifiable actors and actionable remedies. Whether academia can actually mobilize at scale is a separate question, but the duty framework is sound.
