Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.

The analysis, cited by Crypto Briefing and TechCrunch, reports that Grok recommended Republicans in 92 percent of qualifying AI ballot answers. By contrast, GPT’s selections skew toward Democratic candidates, though no exact percentage is given. The study also notes that other chatbots such as ChatGPT, Gemini and Meta AI do not share a uniform political color. Coverage of the findings appears across five technology outlets. Incyber news highlights the divergent partisan signals among ChatGPT, Gemini and Meta AI, underscoring that the models “don’t share the same political color.” NeoTeo situates the work within a broader “Chatbot political bias study” that examined six Swedish models.

TOKENPOST reports the specific statistic on Grok’s Republican recommendations. TechCrunch headlines the story with the phrasing “Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.” The study arrives amid growing scrutiny of algorithmic impartiality in democratic processes. Prior debates have centered on how language models handle politically charged prompts, and regulators have called for transparency in model outputs. By focusing on Swedish‑developed models, the research adds a geographic dimension, suggesting that bias patterns may vary across development ecosystems.

The forced‑choice methodology—requiring the model to select a candidate when presented with a ballot—highlights how underlying training data can translate into overt partisan recommendations. Future reporting is likely to track whether the companies behind Grok and GPT respond with model adjustments or public disclosures. Observers may also watch for additional comparative analyses that expand the sample beyond the six Swedish models, as suggested by the breadth of outlets covering the story. Monitoring statements from the developers and any regulatory inquiries will indicate how the industry addresses the highlighted partisan disparities.