A viral online video claiming to show an artificial intelligence “forecast” of the 2028 US presidential election has drawn fresh attention to the growing use of chatbots and prediction-style models in political commentary, even as analysts caution that the next White House race remains highly uncertain more than two years out from the first primary votes.
The simulation, published by the YouTube channel Election Time, asked Grok, the AI product owned by Elon Musk’s xAI, to generate a full Electoral College map for a hypothetical contest between former Vice President Kamala Harris and the current Vice President JD Vance. The video, which has been shared widely on social media, presents its result as a state-by-state projection of how Americans might vote in 2028, based on a blend of early polling references, betting-market odds and assumptions about recent voting trends.
“In this video, I asked Grok AI to predict the 2028 presidential election and give us a map forecast,” the host says at the start of the segment, before moving through a series of steps intended to mirror how campaign analysts sort states into categories such as “solid”, “likely” and “toss-up”.
The premise is deliberately speculative. While Vance is currently vice president under President Donald Trump, neither party has officially nominated anyone for 2028, and the eventual field could be reshaped by events, policy shifts, economic conditions, court cases, internal party politics and the long, unpredictable grind of an American presidential campaign. Nonetheless, the video leans into early signals to frame its exercise.
On the Democratic side, the host cites what the video describes as early primary polling, placing Harris in front with 32% support, ahead of California Governor Gavin Newsom on 23.8%. It then lists Pete Buttigieg in third at just under 10%, followed by Representative Alexandria Ocasio-Cortez and Pennsylvania Governor Josh Shapiro. The host suggests those numbers reflect a return to form for Harris after her 2024 defeat, saying “many people did write Kla Harris off following her 2024 election defeat”, while adding that newer readings show her “regaining the lead”.
The 2024 election, which returned Trump to the White House, remains the most recent hard data point available for any modelling exercise. Trump and Vance won the Electoral College with 312 electoral votes, while Harris, then the sitting vice president, lost with 226. That outcome included Republican victories across the major battleground states that had decided recent cycles, including Pennsylvania, Michigan, Wisconsin, Georgia, North Carolina, Arizona and Nevada, a set of results that reshaped internal Democratic debates over coalition strength and message discipline.
In the Election Time simulation, another input is betting markets, which the video treats as a rough measure of perceived likelihood that a candidate will enter the race. The host says the odds imply a 56% chance that Harris will run for the Democratic nomination in 2028, compared with what the video describes as an 11.2% figure “just months earlier”. “Today, it is more likely than not that she is going to run again,” the host says, presenting that as a sign of momentum.
On the Republican side, the exercise assumes Vance would be the nominee, and the video describes him as dominating an early field. It cites a figure of 49.2% support for Vance, well ahead of Donald Trump Jr, and lists Senator Marco Rubio on 12.5%, with Florida Governor Ron DeSantis on 9.2%. The host frames the picture in decisive terms: “If nothing big changes, he will very likely become the party’s nominee in the next presidential race.”
From there, the video’s approach is to allocate states by category. It begins with “solid” states, described as those decided by 15 points or more. In the simulation, Vance’s solid Republican column includes large stretches of the Mountain West and Great Plains, alongside much of the South and Appalachia, as well as Ohio, his home state. The host highlights Ohio as a marker of how the political map has moved over time, arguing that with Vance “at the top of the ticket, it is not difficult to see that he will carry the Buckeye State by a solid 15-point margin,” a judgment grounded in how the state has voted in recent cycles.
Harris’s “solid” Democratic states, in the model, mirror the party’s durable strongholds: the West Coast, parts of New England and the Mid-Atlantic, along with Washington, California, Hawaii, Vermont, Massachusetts, Connecticut, Delaware, Maryland, the District of Columbia and Maine’s first district. The video claims the model gives Harris slightly stronger margins in a handful of states than she achieved in 2024, suggesting a small recovery in some traditionally Democratic territory.
After allocating solid states, the simulation gives Vance an early lead, then expands it through “likely” states, defined in the video as margins between five and 15 points. In that category, Grok’s map places Iowa, North Carolina, Florida, Texas, Arizona, Alaska and Maine’s second district on the Republican side. Harris is shown holding “likely” wins in places such as Oregon, Colorado, New Mexico, Illinois, New York, Virginia, Rhode Island and Maine statewide.
The final rounds of the exercise focus on the battlegrounds, where the map is most sensitive to small shifts. In the “lean” category, described as margins between one and five points, the simulation places Nevada and Georgia as leaning Republican, and groups Wisconsin, Michigan and Pennsylvania together as states moving in the same direction. The host notes their tendency to travel as a bloc in modern elections, saying the video emphasises that the three have backed the same candidate in every election since 1992.
At that stage, the model gives Vance a commanding lead, with the map showing him surpassing the 270 electoral votes needed to win. The last suspense, such as it is, comes from two “tilt” states, described as decided by less than one point. The simulation labels Minnesota and New Hampshire as tilt Republican, presenting them as potential late-cycle surprises. Minnesota has not voted Republican in a presidential race since 1972, but the host points to narrowing margins in recent elections and describes it as “essentially a pure toss-up”.
The video ends with a final projection of 326 electoral votes for Vance and 212 for Harris. “This is the 2028 presidential election map,” the host concludes. “According to Grok AI in a matchup between Vice President JD Vance and former Vice President Kla Harris, it gives Vance 326 electoral votes and Harris 212.”
Even within the video’s own framing, the exercise is not a prediction in the traditional sense used by pollsters or election forecasters who build models from vast datasets and update them as new information arrives. It is, instead, a stylised simulation based on a specific scenario, built around a particular assumed matchup and a snapshot of current attitudes and trends as the video’s creator interprets them.
The popularity of the clip reflects a broader public fascination with using AI systems to answer questions that are, by nature, uncertain and highly sensitive to events. Grok, like other large language models, generates outputs based on patterns in the information it has access to and the prompts it receives. While such systems can summarise political landscapes and replicate the language of forecasting, they do not possess knowledge of future events, and they can confidently produce results that look authoritative even when based on debatable assumptions.
For now, the 2028 race remains undefined. Trump’s return to the presidency has already reshaped both parties’ internal calculations about leadership, ideology and succession. On the Republican side, attention has increasingly focused on the role of the vice president and the broader next-generation field. On the Democratic side, the post-2024 period has fuelled debates about whether to rally around familiar national figures or elevate governors and younger lawmakers with regional bases and less political baggage.
In that context, the Grok-based map is best understood as a snapshot of one online creator’s attempt to translate early political signals into an Electoral College picture, rather than a definitive answer about who will win in 2028. The real contest, with its primaries, conventions, televised debates and unpredictable turning points, is still far away.
