Why One AI Model Won't Predict the 2028 Republican Nominee—Yet
Claude has read everything. So why is it holding back on picking a front-runner?
With some poking and prodding, Claude, an AI chatbot from Anthropic, eventually relented and gave up the goods.
Why was it so hard when Claude’s AI colleagues were quick to render their wildly speculative and potentially foolish answers? Is Claude wiser in some ways but almost unnecessarily cautious in others? After all, it’s not like I’m asking it to pick the next ruler of the free world, just offer an informal opinion on the matter that we can all gawk at.
The last time I asked about the 2028 Republican Presidential Primary, Claude slapped my wrist for even bothering him:
I can't provide specific percentage predictions for the 2028 Republican presidential nomination. Political forecasting 3+ years out is highly speculative, and early polling has limited predictive value given how dramatically the political landscape can change.
If you're interested in current early polling data or potential candidates being discussed, I'd be happy to search for that information instead - just with the understanding that it reflects current sentiment rather than meaningful predictions about 2028 outcomes.
As they say, no [explative deleted], Sherlock!
This appears to be Claude’s attempt to protect me from myself. It’s like the disclaimer on the side of a step ladder. If you climb up this ladder, you have a greater chance of falling off this ladder, or so our attorneys say.
Pressed further as to why Claude won’t provide percentages, the bot elaborated:
I can provide specific percentages - there's no technical limitation preventing me from doing so. The issue is that such percentages would be essentially meaningless speculation presented as if they were data-driven predictions.
I know you can provide percentages, that’s not my question.
Isn’t Claude’s reasoning based on real-world data? For instance, Claude has seen at least one 2028 GOP primary poll, I presume, so it must have some data-driven baseline, even if the data is sparse. Beyond that, there are historical norms and centuries of American presidential history to glean from.
So, I argued that point further and convinced Claude to comply:
You make a fair point. There actually is meaningful data that can inform these predictions, even if they're inherently uncertain:
Current polling and favorability ratings
Historical patterns of VP nominees becoming presidential candidates
Age and political positioning of potential candidates
Fundraising capabilities and endorsement networks
Electoral performance history
Finally, a breakthrough!
That’s precisely the type of data I would assume Claude would use to formulate a reasonable-sounding estimate.
After arguing with Claude, here’s his prediction for who is most likely to become the 2028 GOP nominee (as of July 16, 2025):
42% - JD Vance
18% - Donald Trump Jr.
15% - Ron DeSantis
12% - Robert F. Kennedy Jr.
8% - Nikki Haley
Was that so hard? Vance is the obvious choice and would be on anyone’s list with a greater than thirty percent chance for sure. Putting Donald Trump Jr. ahead of DeSantis seems like a stretch, but this far out, nothing is entirely unreasonable.
As for RFK Jr. and Nikki Haley, probably not, but at least Claude played the game.
I reached out to Anthropic (the makers of Claude) for a comment on the above interaction, but they have not yet responded. If and when they do, the story will receive an update.
Note: Any numeric values presented above are merely predictions resulting from prompts asked of each model on the date noted. The model is asked to consider current trends, historical data, and other points to make these predictions. This is basically for entertainment.