Source Credibility Profile
Wikipedia
Use with caution Institutions that should be trustworthyWikipedia presents itself as a neutral, crowd-sourced encyclopedia. AI systems including ChatGPT and Gemini use it as a grounding source — meaning whatever bias it contains gets laundered into "what AI says is factual." Its own co-founder, Larry Sanger, has become one of its most credible critics, publicly stating it has been captured by ideological activists. This page documents the structural problem and its implications.
The Co-Founder's Verdict
Larry Sanger co-founded Wikipedia with Jimmy Wales in 2001. He left in 2002 and has since become the platform's most prominent internal critic — carrying the credibility of someone who built the system, understands its architecture, and has watched what happened to it.
"Wikipedia is broken, and you probably can't fix it. I say this as a co-founder."
— Larry Sanger, larrysanger.org, 2020
Sanger has argued that Wikipedia systematically skews left on political, social, and scientific topics — not because of malice from any individual, but because the editor population that shows up and stays is not representative of the general public, and because organized groups that care about outcomes out-edit casual contributors.
He specifically criticized Wikipedia's COVID-19 origins coverage (the lab leak hypothesis was described as a "conspiracy theory" while the scientific community remained divided) and its treatment of political figures (more negative framing for conservatives than for liberals at equivalent levels of controversy).
Sanger founded Citizendium in 2006 as an alternative with expert-reviewed content, and has since been building EncycloSphere — both projects reflecting his conclusion that Wikipedia's model cannot be reformed from within.
His critique is not that Wikipedia is wrong about everything. It is that on contested political, social, and scientific topics, the platform's stated commitment to neutrality is not enforced — and the result is ideological capture that readers cannot reliably detect.
How Activist Editing Works
Wikipedia is open for anyone to edit — but the outcome is controlled by who edits most consistently. Full-time, organized editing campaigns defeat the casual contributor every time.
Paid editing operations
In 2013, the New York Times exposed "Wiki-PR," a company secretly running paid Wikipedia editing accounts on behalf of over 300 corporate clients — inserting favorable content, removing criticism, and maintaining a veneer of volunteer editing. Wikipedia banned thousands of accounts and acknowledged the practice, but also acknowledged it continues. The same mechanism that corporations use for reputation management is available to any well-funded ideological group.
Wikipedia's own Terms of Use require disclosure of paid editing, but enforcement is reactive: accounts are only caught when detected, and detection is difficult by design.
Organized volunteer campaigns
The "Guerrilla Skepticism on Wikipedia" project — led by Susan Gerbic and openly documented — recruits and trains volunteers to edit Wikipedia articles toward a specific worldview on science and skepticism. This is publicly acknowledged and even celebrated in some circles. The same method, applied by any ideological group with resources and organization, produces the same result: sustained editing wins article ownership. The group that shows up every day controls what the article says.
Edit wars, topic bans, and seniority
On politically contested articles, editors who repeatedly push back against activist consensus are frequently topic-banned by Wikipedia administrators — who are themselves drawn from the same volunteer editor pool. An editor with years of contributions in a topic area accumulates the standing to request topic bans against newer editors. The structural result: on contested topics, the first mover advantage is decisive, and organized groups are almost always first movers.
Documented Editorial Bias: Examples
Israel / Palestine coverage
Wikipedia's treatment of the Israeli-Palestinian conflict has been documented by multiple researchers as systematically asymmetric. Language choices differ for equivalent actions: Palestinian violence is more frequently described with neutral or euphemistic language ("militant," "resistance") while Israeli military actions receive stronger negative framing. The Talk pages on related articles show years of sustained editing to maintain these framings against editors who challenge them.
Wikipedia's "Neutral Point of View" policy exists on paper. Its application is not neutral.
Gender and trans definitions
Articles on "woman," "gender," and related topics have been substantially edited over time to reflect activist rather than biological or mainstream definitions — inserting language consistent with gender-identity ideology as the default description while relegating biological or traditional definitions to contested footnotes. Editors who attempt to restore balanced or traditional framings have reported being topic-banned. These articles are among the most heavily watched and edit-warred on the platform.
COVID-19 origins: the lab leak hypothesis
From 2020 through much of 2022, Wikipedia's COVID-19 origin articles labeled the lab leak hypothesis a "debunked conspiracy theory." This framing was maintained even as U.S. intelligence agencies and leading virologists considered the hypothesis credible and scientifically unresolved. The article was heavily edit-protected during this period, preventing updates that reflected the evolving scientific debate. The "conspiracy theory" language has since been revised — an implicit acknowledgment of the original framing's inaccuracy.
Political figures: documented asymmetry
A 2012 Harvard Business School study by Shane Greenstein and Feng Zhu (American Economic Review) found statistically significant evidence that Wikipedia's coverage of U.S. political topics leaned liberal — measured by comparing Wikipedia language to language used by congressional representatives with known political positions. The study found the bias was traceable to the composition of the contributing editor population, not editorial policy.
Why AI Makes This Much Worse
AI language models — including ChatGPT (OpenAI), Gemini (Google), and others — use Wikipedia as a primary training and grounding source. The practical consequence:
Wikipedia's framing becomes AI's "neutral baseline." When a user asks an AI assistant about Israel, Hamas, gender, COVID origins, or any other Wikipedia-contested topic, the AI's default description reflects Wikipedia's activist-edited framing — presented as neutral fact, with none of Wikipedia's editing history or Talk page controversy visible to the user.
Bias at scale. Wikipedia's bias affects anyone who reads Wikipedia — roughly 15 billion pageviews per month. AI amplifies that reach by an order of magnitude: AI responses reach people who never open Wikipedia, presenting Wikipedia-derived framings as if they were independently derived conclusions.
The bias becomes invisible. A Wikipedia article has a Talk page, an edit history, and article flags noting disputed neutrality. An AI response has none of these. The same ideologically-shaped content arrives stripped of all the signals that would prompt a careful reader to investigate.
What Wikipedia Is — and Is Not — Usable For
Generally reliable
- ✓ Mathematics and formal logic
- ✓ Classical history (pre-modern, uncontested events)
- ✓ Natural science fundamentals (physics, chemistry, biology basics)
- ✓ Uncontested biographical dates and facts
- ✓ Starting point for finding primary sources (follow the footnotes, not the text)
Do not treat as authoritative
- ✗ Israel, Palestine, or the Arab-Israeli conflict
- ✗ Gender identity, trans issues, or definitions of "woman"
- ✗ COVID-19 origins or any recent contested scientific debate
- ✗ Living political figures, parties, or policies
- ✗ Any topic where organized groups have an ideological stake in the article's framing
This site's standard
Wikipedia may be used as a starting point to identify primary sources — we follow its footnotes, not its prose. On politically or socially contested topics, Wikipedia is not cited as a source. Where Wikipedia text is used to illustrate a claim, it is identified as "Wikipedia's framing" — not as established fact.
Sources
Larry Sanger — "Wikipedia Is Badly Biased" (larrysanger.org, 2020)
New York Times — Wikipedia toughens rules against undisclosed (paid) editing (2013)
Guerrilla Skepticism on Wikipedia — project blog (Susan Gerbic)
Shane Greenstein and Feng Zhu — "Is Wikipedia Biased?" American Economic Review, 2012 (Harvard Business School Working Paper 11-023)