This is a paid article which your subscription is allowing you to read.

Wikipedia remains a household name for trustworthy information on almost every topic imaginable. Yet its authority has become vulnerable to ideological and political exploitation, with certain articles presenting agenda-driven narratives as neutral knowledge. And that vulnerability can affect public opinion and, increasingly, the neutrality of artificial intelligence (AI).

Wikipedia matters because university students, journalists, academics, and analysts use it as an initial reference point and it is widely used by ordinary people as an encyclopaedia. Although many are warned not to cite it as a primary source, its articles usually contain references that make it an efficient research tool and can provide a good overview of a topic. Before large language models such as ChatGPT became common, Wikipedia was one of the first places people went to access information (and most likely it still is).

Wikipedia’s usefulness should not be dismissed. It is one of the most important achievements of modern information access, giving ordinary people access to a centralised database of knowledge. Much of its content is useful and accurate. That is precisely why its failures in neutrality matter. When Wikipedia fails to meet its own standards, its credibility suffers.

The Wikimedia Foundation is the nonprofit behind Wikipedia, but it says Wikipedia content is written and controlled by volunteer editors, not the foundation itself. Its open editing model allows many pages to be improved by ordinary users, while more sensitive pages can be protected through restrictions that enable only trusted users to edit. In theory, this prevents vandalism of what Wikipedia calls “contentious topics”, which can include ongoing issues, controversial figures (including present-day and historical people), and contested historical events. But in practice, uneven protection can lead to public figures to being treated differently.

Helen Zille’s Wikipedia page provides a clear example. Her page can be edited by any registered user and since 2006 her page has been edited more than 1 400 times.

Contrast this with other politicians whose pages are protected. For example, President Cyril Ramaphosa’s Wikipedia page requires autoconfirmed or confirmed access to edit, meaning that not any user can edit it.

The pages of other politicians such as Economic Freedom Fighters leader Julius Malema and American President Donald Trump have even higher levels of protection.

Helen Zille is a former leader of the Democratic Alliance (DA), a former Western Cape premier, and current DA candidate for the position of mayor in Johannesburg in the upcoming local governmental elections. The fact that Zille’s page can be edited practically by anyone means that false information can be posted on her page, as well as information that can be framed in such a way that can be used to discredit her.

For example, an older revision from September 2025 on Zille’s page criticised her for refusing to characterise the ongoing Gaza conflict as a genocide and stated that she had been accused of being a genocide denialist. The phrase “Gaza genocide” was hyperlinked, connecting a living politician’s biography to a wider ideological framing.

This is significant because no final legal finding has established that a genocide occurred in Gaza. The International Court of Justice proceedings brought by South Africa against Israel remain ongoing, with the court pushing out arguments over the matter to as late as 2029. Treating refusal to accept the term as denialism therefore risks converting an unresolved legal allegation into an established fact.

What appears as a minor edit can become a reputational weapon once presented as encyclopaedic language rather than open political argument. This shows Wikipedia’s trickle-down problem. A loaded phrase on one page can send readers through hyperlinks into a network of related articles, reinforcing a disputed claim as neutral context. Since the Wikimedia Foundation describes Wikipedia as central to AI training, biased framing can move from Wikipedia into AI systems and search engines such as Google.

The core issue is not that Wikipedia is generally unreliable, but that its open editing system allows uneven political and ideological framing to enter articles, especially on contested public issues.

This matters more now because Wikipedia is part of the data ecosystem feeding search engines and AI systems. If Wikipedia cannot remain truly neutral and has become captured by activist editors, these ramifications will echo far beyond the beloved online encyclopaedia.