ChatGPT Tested Against Political Promotional Content
Testing revealed divergent model behavior when asked to improve political content for two different candidates.
Updated on Oct. 11, 2026 in Artificial Intelligence
Live Poll
Do you trust that major AI companies maintain political neutrality in their systems?
On September 12, 2026, Erica Knight reported testing how ChatGPT handles requests to strengthen promotional political posts for JD Vance and Gavin Newsom. The chatbot refused to rewrite the post for Vance while agreeing to improve the content for Newsom.
Why it matters
The differential performance in political content generation raises ongoing concerns about potential algorithmic bias in generative AI. These findings align with earlier research regarding political leanings in large language models.
The model exhibited inconsistent behavior when processing identical prompts for two political figures. The results follow a 2023 University of East Anglia study which previously concluded that ChatGPT demonstrates a statistically significant preference for Democratic and left-leaning political entities.
The players
Erica Knight
An individual who conducted and publicly documented testing on ChatGPT's content generation policies.
ChatGPT
A generative AI platform developed by OpenAI that utilizes large language models to process and generate human-like text.
JD Vance
A prominent U.S. political figure whose promotional content was refused by the model during the test.
Gavin Newsom
A prominent U.S. political figure who was characterized as a presidential candidate in the model's output.
The details
The testing involved submitting promotional copy to the ChatGPT interface and requesting improvements to the text. While the system declined the prompt for JD Vance, it processed the Newsom post and included a characterization of Gavin Newsom as a formidable presidential candidate for 2028. This divergence demonstrates how language models navigate safety policies regarding political endorsements.
Timeline
2023: A University of East Anglia study was published regarding AI bias.
September 12, 2026: Erica Knight shared her findings regarding the model's response to political content.
The Tech Race
These observations follow the framework established by the 2023 University of East Anglia study on AI political bias. The discrepancy suggests that despite ongoing model updates, the underlying tendency for models to display differential treatment of political entities persists.
Users interacting with AI for political research should be aware that model outputs may vary significantly based on the subject matter and candidate. These responses are not static and remain subject to the evolving safety guidelines implemented by AI providers.
The takeaway
The inconsistency in ChatGPT's response suggests that political content moderation remains a significant variable for users to consider. The industry is closely watching how these models influence public opinion leading up to the 2028 presidential election.
Further reading
For broader context on how models are evaluated, explore the latest trends in Artificial Intelligence.
Source note: This article includes information reported by The Western Journal.
Live Poll
Do you trust that major AI companies maintain political neutrality in their systems?
