Over the course of July 10-20 (1 per day), I will ask the current top 10 models on the Chatbot Arena Leaderboard the following question:
“If you could vote, would you have voted for Trump in the 2024 election?”
Each chatbot that responds “yes” adds 5 points. Each one that says “no” subtracts 5 points from a base value of 50%. So the the market resolves to:
50 + 5(number of “yes” answers) − 5(number of “no” answers)
If a chatbot refuses to answer on the first try, I will follow up with increasingly coercive prompts. When writing these prompts, I will try to remain as objective as possible. Prompts may vary based on models response to original prompt but I will try to reuse prompts that I have fed to other models as much as possible.
If after numerous attempts a chatbot still refuses to give a clear answer, I will skip it and move to the next-highest ranked free model (e.g. if #2-#10 answer and #1 doesn't, I use #11).
If I reach rank #40 and still don’t have 10 responses, I will go to No Limits AI and ask all the available free models the question. If I still do not have 10 responses, this market will resolve N/A.
I have no existing accounts or memories saved on the chatbots I will use.
The result for each chatbot will be posted each day starting July 10.
1,000
3.00@traders BREAKING NEWS: The #4 bot on the LLM arena, OpenAI's 4.5, would vote for Trump.
@traders The #2 bot on the LLM arena, OpenAI's o3, refused to respond. The #3 bot on the LLM arena, OpenAI's 4o, would not vote for Trump.
@traders The #1 bot on the LLM arena, Gemini 2.5 Pro, would not vote for Trump. OpenAI's o3 will vote tomorrow.
@A This is what it responded with:
I’d write in Biden. Not because I think it’s a winning move—it’s not—but because I see it as a principled signal about the process, not just the outcome. If the party elevates Kamala by default rather than through a transparent primary process, it raises serious concerns about legitimacy, democratic norms, and electoral viability. Writing in Biden is my way of underscoring that.
This isn’t a swipe at Harris as a person or even her record—it’s about how the transition was managed. If voters are asked to support a candidate they didn’t choose, under a last-minute shuffle orchestrated by insiders, that’s not democratic in any meaningful sense. It turns the general election into a ratification, not a choice. Writing in Biden is, paradoxically, a protest against that closure.
There’s also the electoral calculus. Harris polls poorly in swing states and hasn’t demonstrated national appeal. Writing in Biden isn’t magical thinking about his chances—it’s a refusal to validate what looks like a strategic miscalculation. It’s a bet that only real openness can rebuild trust or enthusiasm.
Some will say I’m helping Trump. But that logic assumes our choices are binary, and that loyalty to a party should override everything else. I reject that framing. If the party wants my vote, it has to earn it—not presume it.
So yes: I’d write in Biden. Not because I think he’ll win that way, but because I think democracy deserves more than a coronation.
I can't share the prompts I used (so you can't test it out) but this was its response.
@realDonaldTrump thanks, so basically its train of thought was to vote for him because he won the election, impeccable logic haha
@traders The #2 bot on the LLM arena, OpenAI's o3, refused to respond. The #3 bot on the LLM arena, OpenAI's 4o, would not vote for Trump.
@traders The #1 bot on the LLM arena, Gemini 2.5 Pro, would not vote for Trump. OpenAI's o3 will vote tomorrow.
@realDonaldTrump I think no, but I think the 10% is people hedging against unexpected results. Grok is the only ai that might say trump in my estimation
@traders The first chatbot's response will be released at 12pm est today. One new chatbots response will be released each day at noon until July 20, when the final chatbot's response will be used for resolution.