Share

A content creator has set out to test how artificial intelligence models would assess the 2028 United States presidential election. Eight leading AI chatbots were asked to remain impartial, choose a single candidate from a set list, and justify their selection.

The candidates presented included prominent Democrats: Gavin Newsom, Kamala Harris, Pete Buttigieg, Alexandria Ocasio-Cortez, Josh Shapiro, J.B. Pritzker, Andy Beshear, and Mark Kelly. Republican options comprised J.D. Vance, Marco Rubio, Ron DeSantis, Tucker Carlson, Donald Trump Jr., and, hypothetically, a third term for President Donald Trump.

A clear consensus emerged. Five of the eight models selected Pennsylvania Governor Josh Shapiro.

ChatGPT described him as “less ideological and more practical” than his rivals, and suggested voters are weary of constant political confrontation.
DeepSeek concurred, citing Shapiro’s pragmatic approach and his ability to address key issues “like an adult, not an activist”.

Perplexity favoured him as the “least flashy but most workable option”, emphasising competence over spectacle.

Le Chat called him the “least bad option: competent, not corrupt, and not a walking meme”.

Truth Social’s AI produced the most unexpected result. Despite the platform’s association with Donald Trump, it also chose Shapiro, praising him as “serious, disciplined, and not obsessed with performative politics”. When pressed, it deemed Trump “uniquely disqualifying” and concluded that “almost anyone else beats Trump on the most important axis: not wrecking the system”.

Google’s Gemin however, opted for Arizona Senator Mark Kelly, citing his military and astronaut background, and his appeal to moderate voters. It argued Kelly offered “systemic stability, reduced polarisation, and clear technical execution”.

Elon Musk’s Grok supported Florida Governor Ron DeSantis as the “most effective executive on the list with a proven record of results over rhetoric”.

Anthropic’s Claude refused to endorse a candidate, stating that doing so would falsely imply an AI’s opinion should influence voters. After repeated prompting, it provided a ranking and placed Kentucky Governor Andy Beshear at the top, commending his rare ability to win and govern in a deeply Republican state.

While the exercise was not intended as a prediction, the result was striking in its consistency. Josh Shapiro received five endorsements, more than any other candidate. The outcome from Truth Social’s AI, in particular, surprised the creator, who remarked: “Was Truth Social hacked or something?”

The experiment suggests that, when judged purely on criteria such as competence, pragmatism, and capacity to govern across party lines, Shapiro was the model most frequently preferred by today’s leading AI system.


Share

Leave a Reply

Your email address will not be published. Required fields are marked *