Nearly 70,000 Tests Show Most LLMs Lean Libertarian Left
Large language models are increasingly used for search, writing and decision support, raising concerns that consistent political tendencies could shape how information is framed and consumed. AI testing group Unslop.run examined leading systems across the political spectrum to assess whether their responses showed persistent ideological patterns, an issue with implications for model neutrality, public discourse and user trust.
Unslop.run conducted nearly 70,000 political-spectrum tests across 16 major large language models worldwide. All but xAI’s Grok consistently landed in the “libertarian left” quadrant, according to the findings. The analysis attributed the pattern mainly to imbalances in training datasets and to AI systems amplifying views that are already dominant in society.
All Coverage
1 original reportsThe Backstory
The history behind this eventPolitical Bias Test Finds ChatGPT Furthest Left, Grok Only Major AI Model Leaning Right
AI bias research platform Trakkr examined how six major AI models, including ChatGPT, Gemini and Grok, responded to political and social issues. The study assessed whether providers were delivering on their claims of neutrality. Its findings suggest that training data and model design may shape the perspectives through which users receive information on public issues.
Trakkr’s latest test found a gap between the six models’ stated neutrality and their measured positions. OpenAI’s ChatGPT leaned furthest left on economic issues, xAI’s Grok was the only model that leaned right overall, and Google Gemini came closest to political neutrality. The data did not disclose the test date, sample size or monetary amounts, and the findings should therefore be interpreted in light of the question design and scoring methodology.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.