Don’t Trust Grokipedia!!

I periodically search Google for David Crockett Graham to see whether anything new of interest has appeared. Recently, a Grokipedia entry on Graham showed up. If you haven’t heard of it, Grokipedia is a fully AI-generated “alternative” to Wikipedia, produced by Elon Musk’s AI system, Grok.

     At first glance, the Grokipedia entry on Graham appears to have been lifted largely from Wikipedia. As one reads further, however, it seems expanded with additional sections and updates, including references to recent scholarship, even works by Chinese authors. It initially looks impressive. But a closer reading reveals a multitude of errors—AI hallucinations and serious misinterpretations.

     Consider just a few examples. At the end of the article, a section titled “Recent Preservation Efforts” describes the digitization project at Mount Holyoke College and names a “Professor Christopher Crutcher” as its supervisor. There is no such professor. In the preceding section, “Criticisms and Alternative Viewpoints,” the article misconstrues Graham’s disagreement with Thomas Torrance over the origin of the Qiang and the nature of their religion. It was Torrance whose theological commitments shaped his interpretation, leading him to argue that the Qiang were a lost tribe of Israel. Graham rejected that claim and took a more careful approach to the Qiang’s religious beliefs, disputing Torrance’s assertion that they were monotheists. The article also cites one of Bian Simei’s recent publications but refers to her as “Bian Xiang.” It further implies broad disagreement from Daniel Dye, when both Simei’s article and Graham’s autobiography indicate only that Dye sided with Torrance on the question of the Qiang’s origin.

     Returning to the beginning of the article, Grokipedia states that Graham was interned by Communist forces and relocated to the United States, implying prolonged captivity and forced removal. In reality, he was captured by Communist troops in Sichuan in 1933, held for a single day, and released after he quite literally talked his way out. Later, under “Challenges During Political Turmoil,” the article claims he was captured in the mid-1940s. That date is wrong, and the text misses the long timeline of the Communist revolution, going all the way back into the 1920s.

     The section on “Childhood and Family Background” gets his mother’s name entirely wrong. It lists two siblings as influential figures in his upbringing, yet Lewis Israel Graham died before David was five years old. The article implies he grew up in Oregon; in fact, he grew up in Walla Walla, Washington. It emphasizes supposed “hardships of rural life in the post-Civil War South,” portraying him as born into a poor farming family in Green Forest, Arkansas. In truth, his father purchased substantial land and was a successful farmer and stone mason. The family later sold their farm and moved by train to the Northwest, where real hardship struck: rinderpest destroyed their livestock, a broken flume flooded their farmland, tuberculosis claimed his mother, and typhoid took several of his siblings.

     The article also invents details that appear nowhere in Graham’s own writings. Baseball? Graham never mentions playing it. “Specimen collecting” in childhood? He says nothing of the sort. He does describe working long hours in the fields and hunting and fishing to provide food for his family—important formative experiences that Grokipedia does not include.

     This is only a sampling from a brief review of a few sections, yet the errors are pervasive. In this particular case, I can identify them because I have spent fifteen years intensively studying Graham’s life: reading his autobiography multiple times, examining hundreds of his letters, studying his publications, reviewing contemporary newspaper accounts, and working through secondary scholarship. I know the terrain. But Grokipedia contains thousands of articles generated without transparent human editorial oversight. If I were reading about a subject outside my expertise, I might not recognize the errors. The prose sounds confident and authoritative. It would be easy to be misled.

     This raises a broader concern. AI tools can be useful. In a conversational format, if ChatGPT produces an error, one can challenge it and request clarification. Even then, users must exercise judgment. But Grokipedia presents itself as a finished, reference-style product. It is described as open source, yet it is controlled by xAI and Grok. No one can directly edit an entry. Errors may be reported, but the process by which they are evaluated and corrected is opaque.

     Wikipedia has its flaws, but it operates with transparent editorial mechanisms and visible revision histories. Grokipedia does not. Based on my review of its article on David Crockett Graham, I would offer a simple caution: do not trust Grokipedia as a reliable source.