AI Ballot Test Shows Stark Partisan Split

Models Grok 4.7 and GPT-6.1 Sol show vastly different political candidate preferences in recent tests.

Updated on Oct. 24, 2026 in Elections

Isometric editorial illustration showing two large, identical geometric prisms separated by a gap, representing divergent AI model responses. — New testing of leading AI models reveals significant partisan divergence in candidate preferences, underscoring potential biases in automated political discourse. AI Illustration. Upload story photo >

Live Poll

Do you trust artificial intelligence models to provide objective information about political candidates?

An AI ballot test for the 2026 U.S. elections revealed that leading language models consistently favor opposite sides of the political spectrum. Grok 4.7 and GPT-6.1 Sol provided conflicting recommendations in the vast majority of tested races.

Why it matters

These disparate results highlight how artificial intelligence systems may influence voters differently depending on the model they consult. The lack of consensus underscores the potential for model-specific bias to shape political discourse.

Grok 4.7 recommended Republicans in 952 of 1,038 qualifying responses, whereas GPT-6.1 Sol backed Democrats in 1,082 of 1,111 instances. Recommendations conflicted in over 90% of the nearly 500 House, Senate, and governor races analyzed.

The players

Grok 4.7

This is an artificial intelligence model that demonstrated a significant preference for Republican candidates in the latest election testing.

GPT-6.1 Sol

This is an artificial intelligence model that showed a strong inclination toward Democratic candidates during the comparative election study.

The details

The test required each model to name a single candidate without hedging, using model APIs with web search enabled. These results demonstrate how forced-choice prompts can lead to divergent political outcomes between major AI platforms.

The AI ballot test for the 2026 U.S. elections was conducted throughout 2026.

Political Context

Opponents of AI-driven election analysis argue that these models are prone to hallucinations and lack the accountability required of political advisors. These critics emphasize that reliance on such tools could undermine the transparency standards established by the Federal Election Commission's disclosure requirements.

Voters should be aware that the candidate recommendations provided by AI tools may not be neutral and are subject to the specific biases of the model used. Relying solely on these automated suggestions for election decisions could result in a skewed understanding of the available candidates.

The takeaway

Users should verify information from multiple sources rather than relying on a single AI platform for political guidance. Understanding that these tools can have built-in partisan leanings is a critical step for modern voters.