Most major AI chatbots still lean left on political questions, even "anti-woke" models are no exception

Insight summary

• •Most major AI chatbots demonstrate a left-leaning bias on political questions, including those marketed as conservative.
• •OpenAI's GPT-5.5 showed 80% left-leaning answers, while Deepseek's V4 Pro had 70%.
• •xAI's Grok 4.3 produced more right-leaning answers than others but still had a majority of left-leaning responses.
• •Google's Gemini 3.1 Pro was the most balanced, presenting both sides 93% of the time.
• •The bias may stem from training data overlaps and some deliberate interventions on specific topics, like trans rights.
• •The investigation highlights the difficulty of categorizing AI responses simply as left or right due to conflicts with scientific consensus or human rights.
• •Full analysis and code are available on GitHub for transparency and further study.