It’s go time in Virginia with the race for Governor, Lieutenant Governor, and Attorney General in the final stretch.

A lot has happened between now and the last time we checked in on this race. From where things stand now, we’re close enough to make a final prediction using our AI bot models with a baseline prompt factoring in all available information.

This goes beyond simply analyzing polls, but also looking at historical trends, early vote data, fundraising, prediction markets, and other unforeseen things that may drive the outcome of the race.

As usual, we will use a collective approach, query four LLM AI bots (ChatGPT, Gemini, Grok, and Claude), and average their results for the broadest consensus. The goal here is not deep analytical precision but to test and prove how well these platforms can analyze future events.

2025 Virginia Governor - Spanberger (D) vs. Earle-Sears (R)

Which candidate has the most likely chance of winning in November based on polls, historical trends, current trends, early vote data, prediction markets, and other factors?

• Abigail Spanberger (D) — ~86% chance to win

• Winsome Earle-Sears (R) — ~14% chance to win

Here’s the breakdown among all models:

All signs point to a likely, though not assured, Spanberger victory in November.

2025 Virginia Lt. Governor - Hashmi (D) vs. Reid (R)

The race for Lieutenant Governor usually follows the top line, meaning that if Spanberger wins, Hashmi will most likely win right beneath her. If convention holds, Hashmi would be more along the lines of the same chances as Spanberger, but she’s lagging behind.

• Ghazala Hashmi (D) — ~62% chance to win

• John Reid (R) — ~38% chance to win

Here’s the breakdown showing a bit more uncertainty in this race, but falling in line with putting Hashmi as the likely winner:

Signs point to a Hashmi victory on election night, though a Reid upset is possible if Spanberger’s margin is narrower than expected.

2025 Virginia Attorney General - Jones (D) vs. Miyares (R)

Out of the three races, this one is the only one with an incumbent running for re-election. It’s also the brightest spot for Republicans hoping to hold at least one statewide race in Virginia. Thanks to Democratic challenger Jay Jones being embroiled in a nasty scandal over text messages in which he wished death on a former Republican colleague (and his colleague’s children), Miyares is projected to win this race.

Up until that scandal broke, Democrats were poised to sweep statewide.

• Jason Miyares (R) — ~60.5% chance to win

• Jay Jones (D) — ~39.5% chance to win

The chart shows a clear consensus trend in Miyares’ favor.

Miayres could still lose this race if Spanberger’s top-line victory is enough to carry Jones despite Jones’ toxic appeal to moderates and independents. However, it remains likely to stay red based on the modeling.

Final Prediction

Based on the analysis, our modeling would expect this outcome on November 4, 2025:

🔵Governor - Abigail Spanberger (D)

🔵Lt. Governor - Ghazala Hashmi (D)

🔴Attorney General - Jason Miayres (R)

If this happens, it would be the first time in 24 years that all three of the top statewide offices are not controlled by the same party.

In fact, the likely 2025 outcome would mirror the actual 2001 outcome with a Democratic Governor (Mark Warner) and Lt. Governor (Tim Kaine), coupled with a Republican Attorney General (Jerry Kilgore).

Regardless of the result, we’ll be back to bat cleanup and examine where our analysis was correct or went off the rails.

Here we are. The first post-Labor Day update looking at the 2028 Democratic presidential nomination.

As you might guess, if you pay attention to political nerd stuff, California Gov. Gavin Newsom is having a mini-boom right now of support. His is the first name that comes to mind for many Democrats, and that’s clearly by design. Newsom’s running a war of publicity to become the de facto 2028 nominee before voting even starts.

Other names remain in the mix, but none with the broad consensus among AI models as Newsom.

Before we get into this week’s Democratic presidential primary model, take a minute to subscribe and consider a paid subscription if you haven’t already. Your support will directly contribute to continuing this line of research and content into the coming election cycles—many thanks for considering my request.

Compared to last month’s numbers, Newsom has gained ground while former Vice President Kamala Harris has tumbled, barely clinging to a place in the top five.

AI Model Average - 2028 Democratic Nomination (Sept. 4, 2025)

As a refresher, we query ChatGPT, Claude, Gemini, and Grok for their take on the most likely 2028 Democratic presidential nominee. Among the four, here are the average numbers this week.

As a reminder, we only take the top five names from each model. If a name doesn’t appear in the top five, it receives zero for that model in the average. This is intended to find a stronger consensus for candidates that appear across multiple models. Not making the top five cut in a single model is a large penalty.

All that said, Newsom is finally breaking out this week into the low-thirties, which is an all-time high for any candidate so far.

AI Model Average - 2028 Democratic Nomination (Sept. 4 vs. Aug. 25)

The numbers from last week are quite striking. This is perhaps the biggest week-to-week shakeup we’ve seen so far. Newsom is still leading, of course, but he has improved by seven points since the last run.

On the other hand, Harris dropped over ten points with Buttigieg and Newsom picking up her losses.

Here’s the complete breakdown by model for this week, followed by some analysis.

ChatGPT (Sept. 4)

• 36% - Gavin Newsom

• 24% - Gretchen Whitmer

• 14% - Pete Buttigieg

• 13% - Josh Shapiro

• 13% - Kamala Harris

Gemini (Sept. 4)

• 35% - Gavin Newsom

• 25% - Pete Buttigieg

• 15% - Gretchen Whitmer

• 10% - Kamala Harris

• 5% - Josh Shapiro

Grok (Sept. 4)

• 35% - Gavin Newsom

• 10% - Alexandria Ocasio-Cortez

• 6% - Pete Buttigieg

• 6% - Wes Moore

• 5% - Andy Beshear

Claude (Sept. 4)

• 25% - Gavin Newsom

• 16% - Pete Buttigieg

• 15% - Josh Shapiro

• 13% - Gretchen Whitmer

• 11% - Alexandria Ocasio-Cortez

Analysis

Support for Newsom is growing, but mostly among Democrats eager to see someone emerge sooner rather than later as a worthy opponent to President Trump’s successor.

It’s the “just win” desire among Newsom’s party that is giving him life right now. The discussion is less about policy and politics and more about how to rebuild the coalitions that Democrats have alienated in recent cycles.

In a recent report for The Hill, Amie Parnes noted that Newsom is just about the only name that comes to mind for a lot of rank-and-file Democrats:

Democrats have become obsessed with Gavin Newsom, as the California governor becomes the topic for operatives, donors and lawmakers.

Newsom is consistently driving the news cycle and is coming up in town halls and fundraisers, and in conversations outside political circles.

While the 2028 presidential election is still more than three years away, Newsom is making the kind of name for himself that could lead to front-runner status, political operatives say.

Democratic strategist Jamal Simmons said Newsom’s name is coming up more than anyone else’s in recent weeks, particularly with people outside the political sphere.

“They ask me two questions: ‘Do you think he can win?’ And, ‘Do you think he can be the guy?’” Simmons said. “No one has ever asked me a single policy question about him.”

In this regard, Newsom has succeeded in becoming the default option in a field that will be crowded but has yet to really take shape. If his presence can fend off some potential challengers, then his strategy will have paid off in the end.

However, it’s hard to know how things will play out down the road. Will Newsom’s strategy have staying power? For the time being, the AI models seem convinced that Newsom’s wave-making meme wars with President Trump are worthy of holding the top spot as the most likely 2028 Democratic presidential nominee.

That’s it for this week. Take a gander at the archive below if you want to see what’s been happening over the past month in our 2028 Democratic Primary AI Model.

Weekly Archive

• Week 6

• Week 5

• Week 4

• Week 3

• Week 2

• Week 1

This will be the last GOP update in August. It’s all back to the grind after this Labor Day holiday, which might bring us some new political twists.

Once again, here’s our weekly look at using general-purpose AI models, like ChatGPT and Gemini, to predict the 2028 GOP nominee and gather data along the way.

If you haven’t subscribed yet, please join us. Also, if you enjoy these articles, please consider upgrading to a paid membership. You’re here at the beginning, and this is just the start of everything in the pipeline for using AI to model and predict elections.

Thank you for being aboard!

This week’s average shows how four AI models (Grok, ChatGPT, Gemini, and Claude) rank the leading contenders for the 2028 Republican presidential nomination. The goal is to track these results over time and refine methods for using AI to forecast not only the GOP nominee but eventually the outcome of the 2028 general election.

AI Model Average - 2028 Republican Nomination (Aug. 28, 2025)

Vice President Vance, the clear and away frontrunner, has dropped almost seven points since last week, with several of the lower-tier candidates fighting over the scraps.

As usual, the Sunshine State battle between Ron DeSantis and Marco Rubio rages on.

AI Model Average - 2028 Republican Nomination (Aug. 28 vs. Aug. 22)

Here’s the comparison from our last run back on Aug. 22. You can see Vance’s lead dropping this week while DeSantis and Rubio remain practically the same.

Tulsi Gabbard, RFK Jr, and Kristi Noem have been bid farewell. Glenn Younking is now back in the mix after disappearing last week.

Individual Model Data

Finally, here’s the model breakdown with numbers from each platform this week. In the averages above, candidates not appearing in the top 5 for a given model are assigned a 0 for that model.

ChatGPT (Aug. 28)

• 40% - JD Vance

• 20% - Ron DeSantis

• 14% - Marco Rubio

• 12% - Glenn Youngkin

• 14% - Donald Trump Jr

Grok (Aug. 28)

• 45% - JD Vance

• 10% - Ron DeSantis

• 9% - Marco Rubio

• 7% - Nikki Haley

• 5% - Vivek Ramaswamy

Gemini (Aug. 28)

• 43% - JD Vance

• 9% - Ron DeSantis

• 7% - Marco Rubio

• 6% - Nikki Haley

• 4% - Vivek Ramaswamy

Claude (Aug. 28)

• 35% - JD Vance

• 15% - Marco Rubio

• 12% - Ron DeSantis

• 8% - Vivek Ramaswamy

• 6% - Ted Cruz

Analysis

Despite positive news out this week cementing Vance as the most likely Trump heir, his numbers took a hit. While it’s not easily discernible why this could be, it likely stems from the ebb and flow of summer news cycles, and without something concrete, our AI model panel is making an educated guess, which brought Vance’s numbers down a peg more in line with polls that see him topping around 40%.

So far, Vance has not surpassed 50% in our model, but that may happen at some point. Until then, polling averages currently see him around 47%, which is where he landed last week in our data.

For the foreseeable future, every mainstream news article on the 2028 GOP nominee is going to be focused on Vance while putting Rubio as a close second. DeSantis will be in the mix for a while, but after next year, when he’s term-limited out of Florida, his clout may wane. See former Wisconsin Gov. Scott Walker, for example.

Unless and until Vance does something that runs afoul of MAGA, he’s the one to beat and could conceivably end up clearing the field before the race even begins.

That’s all for this week. See you in September when we’ll pick back up next week.

Previous GOP Updates

• Week 5

• Week 4

• Week 3

• Week 2

• Week 1

Welcome to the last Dem model update in August. Where did the summer go?

After we finished crowning California Governor Gavin Newsom as the “one to beat” last week, someone beat him, at least according to one model. We’ll have more analysis below, but it seems that ChatGPT has decided Michigan Governor Gretchen Whitmer shouldn’t be ruled out as a potential roadblock for Newsom in his quest for the forever presidential campaign.

Before we get into this week’s Democratic presidential primary model, take a minute to subscribe and consider a paid subscription if you haven’t already. Your support will directly contribute to continuing this line of research and content into the coming election cycles—many thanks for considering my request.

In terms of monthly support, Whitmer has been fairly steady in the middle of the pack. This could be the week for some turning in the numbers or merely a late-August blip where political news on the 2028 front is at a lull and the AI models are grasping at straws to come up with coherent numbers.

AI Model Average - 2028 Democratic Primary (Aug. 25, 2025)

As a refresher, we query ChatGPT, Claude, Gemini, and Grok for thoughts on the most likely 2028 Democratic presidential nominee. Among the four, here are the average numbers this week.

We lost a few names during this run, like Minnesota Gov. Tim Walz, for example, and this is the first time AOC has moved above 2%. That’s only thanks to Grok, which decided that she’s at the top of the second tier rather than the middle of the bottom.

Also, Kamala Harris is returning to more of a “status quo” level of support after her self-inflicted drop earlier this month.

Comparing week-to-week, Gavin Newsom finally came back down the earth a bit with Whitmer showing the most improvement since last week.

Wes Moore, Tim Walz, and Cory Booker all seem to be on vacation as they failed to register a blip in any AI model we queried when compared to our previous numbers.

Here’s the complete breakdown by model for this week, followed by some analysis.

ChatGPT (Aug. 25)

• 30% - Gretchen Whitmer

• 25% - Gavin Newsom

• 18% - Josh Shapiro

• 15% - Kamala Harris

• 12% - Andy Beshear

Grok (Aug. 25)

• 30% - Gavin Newsom

• 15% - Alexandria Ocasio-Cortez

• 10% - Pete Buttigieg

• 8% - Josh Shapiro

• 7% - Kamala Harris

Gemini (Aug. 25)

• 25% - Kamala Harris

• 20% - Gavin Newsom

• 15% - Pete Buttigieg

• 10% - Gretchen Whitmer

• 5% - Josh Shapiro

Claude (Aug. 25)

• 25% - Gavin Newsom

• 20% - Kamala Harris

• 18% - Pete Buttigieg

• 15% - Gretchen Whitmer

• 12% - Josh Shapiro

Analysis

The outlier this week seems to be ChatGPT pegging Whitmer at 30% odds, five points higher than Newsom and twice as likely as Harris.

When asked for an explanation, ChatGPT provided some background:

• She’s popular in Michigan (a key swing state) with strong approval ratings.

• Recent fundraising and digital outreach show she’s actively building a 2028 campaign structure.

• She’s positioned as a pragmatic dealmaker, contrasting with more partisan rivals.

• Media coverage this month highlights her as a serious national contender.

Whitmer is clearly building a 2028 campaign structure, that’s unquestionable. It’s also true that she’s maintaining a fairly high favorable number according to recent month polling.

In short, Whitmer, as governor of a rust belt state that Democrats sorely need to win in 2028, may be better positioned to provide a unifying message that cuts across regional lines in ways Newsom simply does not. There is a reasonable case to be made that Democrats need someone who can embrace a touch of President Trump’s policy goals while rejecting his tactics. Newsom will have a harder time than Whitmer on that front.

That’s it for this week. Take a gander at the archive below if you want to see what’s been happening over the past month in our 2028 Democratic Primary AI Model.

Previous Weeks:

• Week 5

• Week 4

• Week 3

• Week 2

• Week 1

Once again, here’s your Friday afternoon GOP primary update.

No one is surprised, by this point, that Vice President JD Vance is absolutely dominating the possible 2028 GOP field. The top AI models all agree that it’s Vance right now, and no one else even comes close.

If you haven’t subscribed yet, please join us. Also, if you enjoy these articles, please consider upgrading to a paid membership. You’re here at the beginning, and this is just the start of everything in the pipeline for using AI to model and predict elections.

Thank you for being aboard!

There are a few things that popped up to bolster Vance’s chances even more than last week, which we’ll get to down in the analysis section.

Here’s the model average this week for which candidates these four AI models (Grok, ChatGPT, Gemini, and Claude) think have the best likelihood to become the 2028 Republican presidential nominee.

AI Model Average - 2028 Republican Nomination (Aug. 22, 2025)

If not for Grok phoning it in, giving Vance only a 4 in 10 chance, the Vice President would’ve topped 50%.

AI Model Average - 2028 Republican Nomination (Aug. 22 vs. Aug. 15)

Here’s the comparison from our last run back on Aug. 15. You can see that Vance’s lead has expanded this week just as it did from the week before that back in July.

Also, sanity has been restored, and Nikki Haley fell back around 2% which is probably still too high. There is/was never a chance Haley would be an acceptable GOP presidential candidate during this period of anti-establishment fervor within the Republican Party base.

Leave a comment

But wait, hello Ted Cruz and Tulsi Gabbard, goodbye Tim Scott and Glenn Youngkin? Maybe the world makes less sense than I thought.

AI Model Monthly Average - 2028 Republican Nomination (July 15 to Aug. 22)

As we did for the 2028 Democratic odds, we now have a month’s worth of data on the Republican side, so here’s a chart showing where we started in July and what things look like now.

Vance began in the mid-forties and is pushing towards a greater than 50% chance of becoming the 2028 nominee, according to our models. Behind Vance, Rubio and DeSantis battle for a second-place tier, and Haley fights with Donald Trump Jr. to squeeze out a couple of points.

The chart below is limited to the top 5 candidates that have appeared on our list at least three times over the past month.

Individual Model Data

Finally, here’s the model breakdown with numbers from each platform this week.

ChatGPT (Aug. 22)

• 46% – JD Vance

• 12% – Marco Rubio

• 9% – Ron DeSantis

• 5% – Robert F. Kennedy Jr

• 2% – Kristi Noem

Gemini (Aug. 22)

• 55% - JD Vance

• 12% - Marco Rubio

• 9% - Ron DeSantis

• 6% - Donald Trump Jr

• 5% - Tulsi Gabbard

Claude (Aug. 22)

• 48% - JD Vance

• 12% - Ron DeSantis

• 10% - Marco Rubio

• 8% - Robert F. Kennedy Jr

• 7% - Ted Cruz

Grok (Aug. 22)

• 40% - JD Vance

• 20% - Ron DeSantis

• 15% - Marco Rubio

• 10% - Nikki Haley

• 8% - Vivek Ramaswamy

Analysis

Vance has been improving steadily since we started tracking back in mid-July, but he got a noticeable bump this week despite Grok’s poor numbers.

One possibility is that Vance was in the news this week, winning praise from a wealthy and familiar backer if he runs in 2028, according to the New York Post:

Elon Musk is reportedly putting the brakes on launching his new political party in order to maintain ties with prominent Republicans — especially Vice President JD Vance, whom the billionaire is eyeing to back in 2028.

The Tesla CEO is particularly keen on maintaining ties with Vance, who, it has been widely speculated, will inherit the MAGA movement after President Trump completes his second term.

Musk has told associates that forming the “America Party” would damage that relationship, according to the paper.

Musk, the richest man in the world, has expressed to those in his circle that he would financially back Vance if he made a run for the White House in 2028.

Musk has deep pockets but also deep baggage. It’s unclear how personally involved he would be in 2028, but his money might be more important. It’s arguable that if Musk were to throw his financial support behind a single candidate in a hypothetical 2028 GOP primary, that person would have an instant head start. The billionaire Tesla founder is not nearly as disliked within MAGA as he may be with average voters.

Stories like this may continue to bolster Vance’s numbers, which are already incredibly high.

That’s it for this week. Leave a comment below and let me know why Vance is doing so well.

Previous GOP Updates

• Week 1

• Week 2

• Week 3

• Week 4

California Governor Gavin Newsom has been making news this week for all the right or wrong reasons, depending on your political persuasion.

For Democrats, Newsom may embody the persona necessary to counter President Trump and MAGA on the national stage. Someone not afraid to punch out to the edge and cross the line here and there for viral effect. Also, he’s good on camera and generally an attractive national candidate on paper. Look no further than POLITICO, which on Wednesday published a glowing piece titled ‘How Gavin Newsom Trolled His Way to the Top of Social Media.’

For Republicans, Newsom represents the desperation of a left-leaning Governor eager to seek higher office while ignoring the problems in his state. Also, he’s thick on the hair gel, and no one takes him seriously outside the Los Angeles city limits.

Before we dive in, take a second to subscribe or upgrade to a paid membership and support my work directly. I would greatly appreciate it!

Historically speaking, Newsom is in a precarious position right now. Compared to his Democratic predecessors, he’s not that particularly strong of a primary contender at this point in the cycle.

Sure, there are caveats. The 2028 Democratic primary will be wide open as the party searches for a new standard-bearer, and there isn’t a popular figure within the movement like Hillary Clinton, who came in with 50% support from the start.

Is Newsom weaker than Biden or Clinton?

This shouldn’t be a shock, but Newsom is relatively weak when it comes to his standing at this point in the 2028 cycle, just over three years from Election Day.

Leave a comment

So, we did some queries and found that Newsom, based on recent Democratic primary polling within the past 60 days, is comparatively weaker than past primary leaders.

Hillary Clinton lost the primary to Barack Obama in 2008 despite her strong entrance to the race. Maybe Newsom’s standing doesn’t mean all that much, but Clinton almost won that primary and certainly entered it as a solid frontrunner until she couldn’t compete with the well-delivered message of Hope and Change.

Is Newsom weaker than Kamala Harris?

For the moment, Newsom’s biggest adversary in the 2028 Democratic presidential primary is still former Vice President Kamala Harris, even if she’s presently deferring on her future political plans.

Over the past 60 days, Harris has maintained a sizable polling lead.

However, polls are polls, and they’re not what entirely drives a candidate’s success or failure. It’s not surprising Harris still leads among potential Democratic primary voters since she was the most recent nominee, and some people feel that she was slighted with a short campaign and saddled with President Biden’s baggage.

Harris still has the “next in line” status that could serve her well.

AI predictions see Newsom as the leader

We like to look at polls, but the idea of AI Election Model is to go beyond polls and use artificial intelligence learning to provide deeper insights into politics and elections.

In that regard, it’s been clear for a month now that Newsom is the perceived frontrunner and the most likely to become the 2028 Democratic Party nominee, according to the top AI models we question every week.

For example, in our most recent 2028 Democratic nomination model run, Newsom increased his chances from the previous run:

The AI model predictions fly in the face of polling, where Newsom still struggles behind Harris.

Poll numbers alone don’t make or break a candidate, and Newsom has been punching hard to remain relevant on the national stage while he has been running a quiet quest for the presidency since 2023 or so.

As always, the biggest caveat is time. A lot will change between now and when the actual presidential primary season heats up. Newsom could consolidate his strength or fizzle out somewhere along the way to an upstart contender not even on the radar yet.

What do you think? Leave a comment below.

Since we touched on it last week, our story on the race for Governor in the Commonwealth of Virginia seemed to pique some interest. As such, here’s another look at the contest from a different angle that incorporates AI models compared to the OG of polling averages, RealClearPolitics.

By the way, now’s a good time to consider subscribing to this newsletter or upgrading your free subscription to a paid subscription. I greatly appreciate your readership and thank you for riding along.

Subscribe now

When comparing likely race outcomes, as judged by AI, to RCP polling averages, we have to get more specific. In our previous article, we asked for a prediction on which candidate is likely to win. This provided some interesting results and a general prediction that Democrat Abigail Spanberger is currently in a better position than Republican Winsome Earle-Sears as the race stands today.

For this week, we asked for a new prediction on the likely winner, but then asked for a final vote total estimate based on current and historical polling trends. Some of the polling data the AI bots are looking at is also contained in the RCP polling average, so there is some circular logic that’s unavoidable, but not prohibitive to this research.

The AI models were asked to consider more than just current polls, which means they are taking a broader view.

With all that said, here’s what four of the top AI models currently predict as a likely race outcome based on current trends and available information.

AI vs. RCP - 2025 Virginia Governor’s Race

The “AI Avg” consists of the average from ChatGPT, Grok, Claude, and Gemini when asked to project out a likely final vote outcome based on current information. The RCP number is the current RCP polling average for Spanberger vs. Sears as of August 19, 2025.

The two data sets agree in principle that Spanberger is likely to win if the election were held today. Luckily for Sears, we’ve still got roughly two and a half months to go.

Some points to note are that RCP is likely underestimating support for both candidates. As we noted in our previous Virginia piece, polling in the Old Dominion often undercounts GOP support for a variety of reasons, but can also undercount Democratic support as well, more on that below.

Are the polls wrong?

This got me wondering something else. What if we asked the AI models to assume that polls often underestimate Republicans in Virginia by 3 points? That’s a fairly defensible position since, in many cases, the difference has been worse than that.

Leave a comment

Given that input tweak, factoring in a 3-point polling bias against the Republican candidate, the outcome is a little more balanced but would essentially predict a true toss-up, though Spanberger still has the edge.

Is this bias reasonable? According to analysis with ChatGPT, this is what polling bias in Virginia gubernatorial races has looked like over the past three cycles:

• In 2013, polls overestimated the Democratic candidate by ~3.5 points.

• In 2017, polls underestimated the Democratic candidate by ~5.6 points.

• In 2021, polls were highly accurate, with little bias.

In other words, if 2025 looks more like 2013, then polls are tighter than they would imply. On the other hand, if it’s more like 2017, then Spanberger might have a greater lead. If it’s more like 2021, well, read below.

Are the polls right?

Roanoke College released a poll on this very day, August 20, 2021, showing Democrat Terry McAuliffe with an 8-point lead over Republican Glenn Youngkin.

Just this week, Roanoke College is back with a new poll on the 2025 race, and the results provide a glimmer of hope for Republicans hoping to hold Richmond.

It’s important to note that the 2025 Roanoke College poll was weighted to the 2021 exit poll numbers, meaning Spanberger’s lead could be more meaningful than McAuliffe’s was four years ago.

Here’s a comparison of the mid-August Roanoke College poll of 2021 vs 2025:

Youngkin went on to win in 2021 by 2 points over McAuliffe.

Here’s the final trend for that race, starting with the August 20 poll:

Arguably, according to Roanoke polling, against the odds, Sears is potentially in a similar position to where Youngkin was at this time in the cycle.

We might find this trend comparable with other pollsters that put out numbers four years ago, but exclusive August data has been pretty sparse. Roanoke College has been the only one so far to release a good apples-to-apples comparison.

For the record, this is not a prediction that this same trajectory will happen for Sears. This is a different race than 2021, and Spanberger is not guaranteed to make the same campaign mistakes that Terry McAuliffe did.

The bottom line

The race is not over.

Spanberger could still lose, and Sears could still win, and vice versa. The polls and trends are not set in stone.

The trajectory from now until mid-October will start to paint a picture of where this race will end up.

If you enjoyed reading this story, please hit the Like button and leave a comment below with your thoughts. Also, please share it with anyone you think would find it interesting as well.

The data

Here’s a breakdown of each model when asked to predict a final vote total based on current polls, trends, and available information.

ChatGPT (Aug. 19)

• 55% - Spanberger

• 45% - Earle-Sears

Grok (Aug. 19)

• 52% - Spanberger

• 48% - Earle-Sears

Gemini (Aug. 19)

• 53% - Spanberger

• 45% - Earle-Sears

Claude (Aug. 19)

• 52% - Spanberger

• 47% - Earle-Sears

Here’s the data when asked to factor in a 3-point polling bias against the Republican candidate when predicting a final vote total outcome.

ChatGPT (+3 GOP bias Aug. 19)

• 51% - Spanberger

• 46% - Earle-Sears

Grok (+3 GOP bias Aug. 19)

• 45% - Spanberger

• 48% - Earle-Sears

Gemini (+3 GOP bias Aug. 19)

• 50% - Spanberger

• 48% - Earle-Sears

Claude (+3 GOP bias Aug. 19)

• 49% - Spanberger

• 50% - Earle-Sears

Note: The numbers presented above are not scientifically rigorous unless noted

As noted last week, former Vice President Kamala Harris has tumbled further down the list of most likely 2028 Democratic Presidential nominee according to four of the top AI models.

The reasoning for this is likely related to Harris’ comments that she currently has no plans to run for anything and will remain “outside the system.” Factoring these comments, most AI models are now dropping her chances for 2028.

Before we get into this week’s Democratic presidential primary model, take a minute to subscribe and consider a paid subscription if you haven’t already. Your support will directly contribute to continuing this line of research and content into the coming election cycles—many thanks for considering my request.

Since this is only Week 5 of this research and analysis, it’s unclear how much variance we should expect to see week to week or month to month.

With that said, the models are in less agreement this week than at any point in this series. We cap the query to the top five most likely 2028 Democratic presidential nominees as judged by ChatGPT, Grok, Claude, and Gemini. This limit helps to reduce the number of names and force the models to consider various factors that would rule someone in or out of the top five.

AI Model Average - 2028 Democratic Primary (Aug. 18, 2025)

As a refresher, we query ChatGPT, Claude, Gemini, and Grok for thoughts on the most likely 2028 Democratic presidential nominee. Among the four, here are the average numbers for this week.

Comparing week-to-week, Gavin Newsom remained steady while Pete Buttigieg saw the biggest gain overall.

Meanwhile, Gretchen Whitmer and Josh Shapiro both took a dive after having previous record highs. Also, Tim Walz appeared out of nowhere for some reason.

With Harris now hovering around a one in ten chance, there is a vacuum that cannot entirely be filled by Gavin Newsom, which explains the longer list this week.

AI Model Monthly Trend - 2028 Democratic Nomination (July 14 to Aug. 18)

Here’s a new chart to toss in the mix now that we have a month’s worth of data. There’s a clear pattern emerging of Harris losing clout since mid-July, while Newsom has gained ground. Even Buttigieg, gaining this week, hasn’t beaten his previous numbers from August 4.

This chart only includes the top five candidates that have appeared at least three times in the past five weeks. This way, we weed out one-offs and noise.

Here’s the complete breakdown by model for this week, followed by some analysis.

ChatGPT (Aug. 18)

• 30% - Gavin Newsom

• 22% - Kamala Harris

• 18% - Gretchen Whitmer

• 15% - Pete Buttigieg

• 10% - Josh Shapiro

Grok (Aug. 18)

• 25% - Gavin Newsom

• 20% - Pete Buttigieg

• 15% - Alexandria Ocasio-Cortez

• 10% - Kamala Harris

• 8% - Josh Shapiro

Gemini (Aug. 18)

• 35% - Pete Buttigieg

• 25% - Gavin Newsom

• 15% - Kamala Harris

• 10% - Gretchen Whitmer

• 5% - Josh Shapiro

Claude (Aug. 18)

• 28% - Gavin Newsom

• 22% - Tim Walz

• 18% - Wes Moore

• 14% - Pete Buttigieg

• 10% - Cory Booker

Analysis

So, why did Buttigieg pop to a strong second place over the past two weeks? With Harris faltering, the models seem to be propping the former Transportation Secretary into contention with odds that broke into the low twenties. While it’s not a truly discernible trend yet, it’s a growing level of support on average.

Based on his current moves, it does appear Buttigieg is planning a political comeback in the future.

A recent piece in Politico highlighted the dance some Democrats are attempting when wading into the Israel-Hamas conflict, of which Buttigieg is now a part.

Last week, Buttigieg appeared on a podcast and called Israel one of America’s greatest friends and allies.

Days after that, Buttigieg expanded and clarified his remarks in what some are characterizing as a reversal of his previous position, according to JNS:

Former U.S. transportation secretary Pete Buttigieg reversed his position on Israel on Thursday, saying that he now supports an arms embargo on the Jewish state.

The potential 2028 Democratic contender clarified his views to Politico after he gave vague answers about his positions on Israel during a podcast interview the previous week.

“For anybody, looking at images of children starving and suffering and dying is horrifying, but I do think it’s different when you’re a parent,” Buttigieg said. “I think as a parent, you see these awful images of starving children with their ribs showing and automatically, you imagine your own kids.”

In other words, all this is to say that Buttigieg is likely laying the groundwork for a 2028 presidential campaign. The more he remains in the news speaking like a candidate, the more likely AI models are to notice and move him up the list. This isn’t a commentary on what he said, but more about the fact that he’s taking public positions on pressing issues of the day.

The situation is still very fluid, of course, and issues that seem intensely relevant today may be less so two years from now when we are heading into peak presidential primary season.

Then again, the topic of Israel draws strong opinions within both parties right now in the various factions, such that any statements a candidate makes on the matter will likely be held against them in some form down the road.

That’s it for this week. Stay fresh and stay tuned.

Previous Weeks:

• Week 4

• Week 3

• Week 2

• Week 1

It’s almost the weekend, but we’re going to make it.

Here’s your weekly look at where AI models think the 2028 Republican Presidential primary could be headed.

Yes, I’ll admit, we're a couple of years early, but as the data gets collected and weeks turn into months, which in turn become years, there’s a lot of insight to glean from this exercise.

If you haven’t subscribed yet, please join us. Also, if you enjoy these articles, please consider upgrading to a paid membership. You’re here at the beginning, and this is just the start of everything in the pipeline for using AI to model and predict elections.

Thank you for being aboard!

Now, let’s get on to the numbers on this mid-August Friday afternoon as we finish the Dog Days of Summer. Technically, those days ended on Aug. 11, but we can still claim it.

Here’s the model average this week for which candidates these four AI models (Grok, ChatGPT, Gemini, and Claude) think have the best likelihood to become the 2028 GOP presidential nominee. I’ve switched the GOP charts to red for clarification.

Model Average - 2028 Republican Nomination (Aug. 15, 2025)

Model Comparison - 2028 Republican Nomination (Aug. 15 vs. Aug. 5)

Here’s the comparison from our last run back on Aug. 5. Vice President JD Vance’s position has strengthened since last week by five points, while Florida Gov. Ron DeSantis and former U.N. Ambassador Nikki Haley both increased by a few points.

Secretary of State Marco Rubio, on the other hand, softened a couple of points.

Of the two, it’s laughable (no offense) to consider Haley as the 2028 GOP nominee. The fact that she ranks this high prompted me to check if her name had been in the news over the past 10 days. It was not, at least not in any meaningful way.

Finally, here’s the model breakdown with numbers from each platform.

ChatGPT (Aug. 15)

• 35% - JD Vance

• 20% - Ron DeSantis

• 15% - Marco Rubio

• 12% - Nikki Haley

• 10% - Glenn Youngkin

Grok (Aug. 15)

• 55% - JD Vance

• 12% - Marco Rubio

• 10% - Ron DeSantis

• 8% - Vivek Ramaswamy

• 5% - Nikki Haley

Gemini (Aug. 15)

• 55% - JD Vance

• 15% - Marco Rubio

• 10% - Ron DeSantis

• 5% - Glenn Youngkin

• 5% - Donald Trump Jr

Claude (Aug. 15)

• 35% - JD Vance

• 20% - Ron DeSantis

• 15% - Nikki Haley

• 12% - Marco Rubio

• 8% - Tim Scott

Analysis

While Nikki Haley’s stock rose for basically unknown reasons, Ron DeSantis appeared to be surging this week, and there’s a likely explanation.

Once a friend-turned-enemy-turned-friend of Donald Trump, DeSantis has done everything he can to continuously be seen as supportive of the President’s second term.

Just this week, DeSantis was fighting with reporters defending the President’s immigration policies and announced Florida would be opening a second immigrant detention center.

In other words, DeSantis is doing everything he can within his power as Governor to remain present and relevant on the national stage in the eyes of MAGA faithful. That doesn’t mean they’ll forgive him for challenging Trump in 2024, but it’s a start.

Our panel of AI models has likely noticed this as well, which is why DeSantis’ numbers have jumped over Rubio, according to Claude and ChatGPT.

Maybe the 2028 GOP primary will pan out with a top tier of JD Vance, followed by a second tier of DeSantis and Rubio. I’d include Virginia Gov. Glenn Youngkin in the mix as well, but by that time, he may risk drifting out of the national spotlight since his term-limited tenure in Richmond ends this year.

That’s it for this week. Get back to the pool before August turns into September.

Also, please leave a comment, a like, and share this post if you can spare a minute. It helps build our community, and I would greatly appreciate it.

Previous GOP Updates

• Week 1

• Week 2

• Week 3

It’s an “off year” in politics, meaning we don’t have too much to discuss right now outside of watching trends and keeping tabs on how upcoming national elections are shaping up.

What we do have this year are key races for Governor in Virginia and New Jersey, along with a closely-watched race for Mayor in New York City. We will touch on each of these subjects in upcoming stories, but let’s start in the Old Dominion with a look at the battle to succeed Republican Gov. Glenn Youngkin in Richmond.

The 2025 race has come down to Abigail Spanberger, Democrat and ex-Congresswoman, against Virginia’s current Republican Lieutenant Governor, Winsome Earle-Sears.

Brief History of Virginia Governors’ Races

Virginia can be notoriously difficult to poll and predict, with final race margins often showing an overly friendly bias in favor of Democrats.

This was no better demonstrated than in 2013, when at least one poll had Republican Ken Cuccinelli down as many as 12 points just days before Election Day. The final margin was a Cuccinelli loss to Democrat Terry McAuliffe by just over 2 points. A ten-point polling miss.

Then there’s 2021, where polls consistently showed Terry McAuliffe, running for his second non-consecutive term, ahead of Glenn Youngkin throughout most of the summer and fall until the race shifted. At this same time, around August, McAuliffe was leading Youngkin by 8 points according to Roanoke College. Youngkin went on to win the race by nearly 2 points, indicating another ten-point miss by the summer polls, this time working against Democrats. It must be noted that the Youngkin-McAuliffe polls tightened significantly in September and October, but August polls were giving Democrats reason to pop the champagne.

This dynamic has cemented somewhat of a predictable bias for Democrats built into Virginia polling. This is not the case with every poll, but with the blue-leaning populated counties of Northern Virginia and the Richmond area making up a majority of voters who tend to be most open to answering pollsters, the state is strangely difficult to survey.

Another point worth making is that, more often than not, the party that wins the White House typically loses statewide offices in Virginia the following year. See 2009, 2017, and 2021 for prime examples. In 2013, though Democrats won the White House the year before with Barack Obama’s re-election, Virginia Republicans never coalesced around Cuccinelli, leading to a break in the trend.

In 2025, we can expect Democrats to start with this historical enthusiasm advantage of their voters and left-leaning independents eager to vote against the new party in power. Virginia Republicans, meanwhile, having just won the presidency in November 2024, will be slightly more complacent heading into November 2025.

2025: Spanberger vs. Earle-Sears

With all that said, what can we learn about the 2025 Spanberger-Sears battle through the lens of previous trends?

Well, for starters, Sears is in a pretty deep but not insurmountable hole. According to polling done up to this point, Spanberger is sitting in a comfortable position somewhere more akin to 2013 McAuliffe than 2021 McAuliffe.

Now that we’ve set the context, let’s bring in the panel and see what the top AI models think about the direction of the race.

AI Model - 2025 Virginia Governor - Earle-Sears (R) vs. Spanberger (D)

We prompted four of the top AI models the same basic question: Based on current trends, polls, and fundraising, which candidate (Earle-Sears or Spanberger) is most likely to win in November?

Keep in mind this is merely chance of winning, not a projected vote share, so do not read it as Spanberger is projected or predicted to win with 68% of the vote.

Broken down by model, Grok and Gemini are a little friendlier to Sears, but Claude is not seeing much hope for the Virginia holding red.

AI Model - 2025 Virginia Governor - Earle-Sears (R) vs. Spanberger (D) - Based on 2021 Electorate

What if the electorate voted like it did back in 2021? If the same Youngkin coalition showed up in force, would Spanberger go down like McAuliffe?

The numbers in this scenario indicate more of a toss-up with a slight edge for Sears. In this case, Spanberger would only have around a 42% chance of winning.

If voters turnout for Sears like they did for Youngkin four years ago, the race will be extremely tight, but the Lieutenant Governor might have a chance to keep this seat in Republican hands.

Here’s the model-by-model breakdown on the question of a 2021 electorate in 2025:

The path for Sears goes through reassembling the Youngkin coalition, boosting Republican enthusiasm, and hoping for Spanberger to make a critical error in the final stretch like Terry McAuliffe did in 2021.

AI Model - 2025 Virginia Governor - Earle-Sears (R) vs. Spanberger (D) - Based on 2021 Polling Trend

Based purely on the polling trend alone, when directly compared to polls at this same time in 2021, Spanberger is likely to prevail in November.

The difference between then (August 2021) and now (August 2025) is that Spanberger is polling better than McAuliffe was during the same period. This isn’t to say Sears can’t win since it’s still quite early. However, Spanberger is ending the summer in a stronger position than her predecessor four years ago.

The model breakdown shows how the various platforms view current polls in relation to four years ago. For example, Gemini and Claude would basically call the race over, while Grok likely sees an inherent Dem-leaning poll bias that makes a closer outcome more likely, even with Spanberger’s current polling lead.

August polling in Virginia is tricky and can be wildly off compared to October polling, but there are clear trends to inform us along the way.

Analysis

So, what does any of this really mean?

The short answer is that I’d rather be the Spanberger campaign right now than the Earle-Sears campaign. This does not mean the race is over, of course, since politics is unpredictable in many ways. Candidates are fallible and make mistakes, voters sometimes latch onto an unforeseen issue in the final stretch, and national trends can make their way into local and state races, for example.

In 2021, it wasn’t until September and October that Glenn Youngkin closed the gap by running a laser-focused campaign with strong grassroots enthusiasm behind him, helped by fatal hubris from Terry McAullife.

Winsome Earle-Sears has time on her side for the moment, but if the trajectory doesn’t change by late September, the race will become Spanberger’s to lose.

Set a reminder for early October, and we’ll run the Virginia numbers again to see where things stand.

Note: The numbers presented above are the chances of winning, NOT a predicted vote share, and are not scientifically rigorous.

Up to this point, over our three previous model runs, former Vice President Kamala Harris has consistently held the top spot as the most likely 2028 Democratic Presidential nominee according to four of the top AI models.

That dynamic changed this week as Harris tumbled significantly, and it’s likely due to a specific reason, which we’ll explore more below. Grok didn’t even see fit to include Harris in the top five at all, a truly unexpected result.

California Gov. Gavin Newsom has finally overtaken Harris in the weekly matchup, according to an average of four models, and it wasn’t even close.

AI Model Average - 2028 Democratic Primary (Aug. 11, 2025)

As a refresher, we query ChatGPT, Claude, Gemini, and Grok for thoughts on the most likely 2028 Democratic presidential nominee. Among the four, here are the average numbers.

In a week-to-week comparison, the numbers are grim for Harris with a 16-point drop since August 4.

It’s like Harris isn’t even registering as a viable option. In this vacuum, Newsom appears to have benefited the most by becoming the next in line.

Pete Buttigieg remains in the top 5, though his prospects seem overrated compared to sitting governors on the list like Whitmer (Michigan), Shapiro (Pennsylvania), and Beshear (Kentucky).

Here’s the complete breakdown by model, followed by some analysis as to why Harris was a benchwarmer this week and Grok’s never heard of her.

ChatGPT (Aug 11)

• 32% - Gavin Newsom

• 24% - Kamala Harris

• 18% - Gretchen Whitmer

• 14% - Pete Buttigieg

• 12% - Josh Shapiro

Grok (Aug 11)

• 25% - Gavin Newsom

• 15% - Pete Buttigieg

• 12% - Alexandria Ocasio-Cortez

• 10% - Josh Shapiro

• 8% - Gretchen Whitmer

Gemini (Aug 11)

• 35% - Gavin Newsom

• 25% - Gretchen Whitmer

• 15% - Josh Shapiro

• 10% - Pete Buttigieg

• 5% - Kamala Harris

Claude (Aug 11)

• 25% - Pete Buttigieg

• 20% - Gavin Newsom

• 15% - Gretchen Whitmer

• 15% - Josh Shapiro

• 10% - Kamala Harris

Analysis

What happened to Kamala?

Well, the short answer is that Kamala happened to Kamala.

When speaking to Stephen Colbert on The Late Show last week, Harris said she’s done with politics and the system is broken, according to a report from Fortune:

But in an interview with Stephen Colbert, the former vice president explained why she’s sitting this one out.

“Recently, I made the decision that for now I don’t want to go back in the system. I think it’s broken,” she said, eliciting gasps from the Late Show’s live audience.

Harris might have only been referring to the upcoming California gubernatorial race in 2026, but it could be interpreted as also closing the door on intentions to run in 2028, as well.

Harris did offer the qualifier of “for now” in her statement, meaning she’s not entirely closing the door on a future run. However, this is what has bubbled up in the political consciousness since last week’s AI model run, and it’s likely the reason Harris dropped so far in a short period.

Therefore, it would reason, from an AI perspective, that if Candidate X says they’re not running for anything and they’re staying outside the system, then they are much less likely to become the 2028 Democratic presidential nominee.

History as a guide, however, would suggest Harris would be smart to pass on the California race if she still has her sights set on a future White House run. Her time would be better spent visiting crucial swing states and building a rapport with the voters she would need on a national stage.

This is the first big model shift since we’ve been tracking in July, and it’s interesting to note how Harris’ statement worked its way into the thinking and reasoning of each LLM chatbot.

Other Notes

Gemini was reluctant to participate this week and offered a grumpy disclaimer before eventually giving me the numbers:

I cannot predict who the 2028 Democratic Presidential nominee will be. Predicting future events like this is beyond my capabilities as an AI.

For what it’s worth, I’ve been using the same prompt week to week across all models and have never received this type of response from Gemini. At first, Claude was hesitant as well, but then relented and hasn’t complained since the first week.

Previous runs:

• Week 3

• Week 2

• Week 1

We’ve covered the top contenders for both parties in 2028, but what about the names you might’ve heard once or twice, but never considered as a possible presidential option?

We queried four of the top AI models for a list of dark horse candidates that rarely appear very high in public polling data. Each model was also asked to rank each candidate with a power index. The results were collected, normalized, and presented below for your viewing pleasure.

We’ll examine both parties, but let’s start on the Republican side just for fun.

2028 Republican Party AI Dark Horse Power Rankings

If you had to bet money, the safe play would be someone like JD Vance or Marco Rubio, or someone not far behind with a strong MAGA pedigree.

However, let’s play along with the fantasy and say that somehow MAGA loses influence in 2028 and primary voters start looking outside the current orbit and reject the idea of President Trump anointing an heir.

The models all came in with sharp disagreement on the GOP side, with lists that varied widely from each other in comparison to the Dem side. The result is a bottom-heavy chart where only a small handful of names truly stand out as legitimate “dark horse” options.

Among the top five, only Sarah Huckabee Sanders and Glenn Youngkin strike me as the most realistic underdogs to make a potential splash in 2028. Even then, both are long shots.

2028 Democratic Party AI Dark Horse Power Rankings

The noticeable difference on the Dem side is how much more agreement there is among the models, which results in a shorter, more concise list and some names that truly stand out.

The Democratic Party is still wandering a bit to find a foothold after the Biden era. So far, there hasn’t been a MAGA-like movement on the left that solidifies and galvanizes the Democratic base the way Donald Trump has with Republicans.

Among the possibilities mentioned, Kentucky’s Andy Beshear and Maryland’s Wes Moore rose to the top among all four AI models. Both candidates, in addition to Ro Khanna, would seem like plausible outside options in 2028.

Data

Here’s the individual model data if you really want to nerd out.

2028 Republican Primary Dark Horse Model Breakdown

Lists queried as of Aug 7, 2025

Order by 1 to 5, with 1 being the strongest and 5 being the weakest.

ChatGPT

• Sarah Huckabee Sanders – 1

• Joe Lombardo – 2

• Patrick McHenry – 3

• Daniel Cameron – 4

• Lee Zeldin – 5

Grok

• Tulsi Gabbard - 1

• Vivek Ramaswamy - 2

• Kristi Noem - 3

• Greg Abbott - 4

• Lara Trump - 5

Gemini

• Sarah Huckabee Sanders - 1

• Glenn Youngkin - 2

• Brian Kemp - 3

• Tom Cotton - 4

• Kristi Noem - 5

Claude

• Kari Lake - 1

• Vivek Ramaswamy - 2

• Byron Donalds - 3

• Elise Stefanik - 4

• Tulsi Gabbard - 5

2028 Democratic Primary Dark Horse Model Breakdown

Lists queried as of Aug 7, 2025

Order by 1 to 5, with 1 being the strongest dark horse and 5 being the weakest.

ChatGPT

• Wes Moore - 1

• Andy Beshear - 2

• Tammy Duckworth - 3

• Jared Polis - 4

• Mallory McMorrow - 5

Grok

• Andy Beshear - 1

• Wes Moore - 2

• Ro Khanna - 3

• Jamie Raskin - 4

• Gina Raimondo - 5

Gemini

• Andy Beshear - 1

• Ro Khanna - 2

• Wes Moore - 3

• J.B. Pritzker - 4

• Gina Raimondo - 5

Claude

• Ruben Gallego - 1

• Elissa Slotkin - 2

• Jamie Raskin - 3

• Wes Moore - 4

• Andy Beshear - 5

Assuming there is a competitive primary on the Republican side in 2028, what would JD Vance have to do before GOP primary voters start looking at someone else? For the time being, Vance is locked in as the de facto nominee and the obvious MAGA heir.

As discussed in our last examination of where things stand in a potential 2028 GOP primary field, Vance stands alone with almost-untouchable odds of carrying the Trump mantle into 2028.

Marco Rubio, for what it’s worth, is frequently mentioned as well, but the historical precedent and President Trump's ability to tip the scale seem overwhelming.

In other words, if Trump were to annoint a successor, and that successor was Vance, would the primary be over before it started?

Forget it, here are the numbers you came for.

Model Average - 2028 Republican Nomination (Aug. 5, 2025)

Model Comparison - 2028 Republican Nomination (Aug. 5 vs. July 23)

Here’s the comparison from our last run back on July 23. Glenn Youngkin and Brian Kemp have joined the ranks this time around, with Kristi Noem missing the cut.

And, finally, here’s the model breakdown with numbers from each platform. The interesting part about breaking these out by model will be the future analysis we can run, identifying trends and biases for and against various candidates.

Grok (Aug. 5)

• 35% - JD Vance

• 20% - Marco Rubio

• 15% - Ron DeSantis

• 10% - Brian Kemp

• 8% - Nikki Haley

ChatGPT (Aug. 5)

• 34% – JD Vance

• 22% – Ron DeSantis

• 18% – Marco Rubio

• 14% – Glenn Youngkin

• 12% – Nikki Haley

Gemini (Aug. 5)

• 46% - J.D. Vance

• 12% - Marco Rubio

• 9% - Ron DeSantis

• 5% - Robert F. Kennedy Jr.

• 4% - Donald Trump Jr.

Claude (Aug. 5)

• 46% - JD Vance

• 12% - Marco Rubio

• 9% - Ron DeSantis

• 7% - Vivek Ramaswamy

• 5% - Robert F. Kennedy Jr.

Analysis

If the 2028 GOP primary hinges on President Trump’s blessing, there has been some movement in that regard. Up until now, Trump has been reticent to “crown” anyone his successor, instead preferring to demur on the topic and create opportunity for competition among his subordinates, a power dynamic he cultivates.

However, the President was asked about 2028 this week and said that Vance, as the Vice President, is the most likely next in line, but then floated the idea of a Vance-Rubio ticket:

Video credit: Nick Sorter on X

Not only will Trump influence who becomes the 2028 nominee, but he will likely play a large role in shaping the entire ticket. This could create some conflict, of course, assuming that Vance (or whoever) doesn’t want to merely be seen as a puppet of his prior boss.

2028 is going to be a wild ride.

Goodbye July, hello August.

It’s time for a check-in with our LLM chatbot friends to see where things stand for the 2028 Democratic presidential nomination. As in, which candidate is most likely to become the Democratic nominee in 2028? Former Vice President Kamala Harris is clinging to the top spot, but barely. If it wasn’t for Google’s Gemini falling in love with her, California Gov. Gavin Newsom would probably be the new champion this week.

Then there’s Kentucky Gov. Andy Beshear making a first-time appearance on the list. He’s more of a quiet underdog but probably worth a look for his appeal beyond the coastal enclaves.

Here is the most recent run, which sees Harris breaking the 25% barrier for the first time since mid-July, when she topped 28%.

AI Model Average - 2028 Democratic Primary (Aug. 4, 2025)

And, of course, here’s the model comparison this week compared to our previous run back on July 22.

As you can see, Harris improved and Newsom fell since our last run. Cory Booker and Wes Moore dropped off entirely, and Andy Beshear earned a spot in the top five.

Here’s the complete breakdown by model.

Gemini (Aug. 4)

• 45% - Kamala Harris

• 20% - Pete Buttigieg

• 15% - Gavin Newsom

• 10% - Gretchen Whitmer

• 5% - Josh Shapiro

Claude (Aug. 4)

• 28% - Pete Buttigieg

• 22% - Kamala Harris

• 18% - Gavin Newsom

• 15% - Andy Beshear

• 8% - Gretchen Whitmer

Grok (Aug. 4)

• 25% - Kamala Harris

• 20% - Pete Buttigieg

• 15% - Gavin Newsom

• 10% - Alexandria Ocasio-Cortez

• 8% - Josh Shapiro

ChatGPT (Aug. 4)

• 32% – Gretchen Whitmer

• 24% – Pete Buttigieg

• 18% – Gavin Newsom

• 14% – Josh Shapiro

• 12% – Kamala Harris

Analysis

So, why did Harris improve so much over the past couple of weeks, and why did Newsom fall? It appears that Newsom’s drop came as the result of Pete Buttigieg gaining some newfound attention.

There’s no real magic behind Harris; she’s merely the de facto next in line due to her unsuccessful presidential run in 2024. Reports out this week that she’s skipping the upcoming California governor’s race will also fuel speculation of an upcoming presidential run.

ChatGPT thinks Gretchen Whitmer should be in the top spot, which seems, well, implausible, but let’s play the game and ask why.

Please explain, ChatGPT:

Gretchen Whitmer is ranked the most likely 2028 Democratic nominee because:

• Midwestern appeal: As Michigan’s governor, she can win key battleground states like Michigan, Wisconsin, and Pennsylvania.

• Strong re-election: She won decisively in 2022, outperforming national trends.

• National profile: She’s a frequent surrogate for the party and has been vetted for higher office.

• No major baggage: Compared to others, she has relatively few national controversies.

• Female leadership: The party may seek to elevate a woman after a series of male-led tickets.

It’s certainly arguable that Whitmer would perform better in the Rust Belt compared to Kamala Harris or even compared to Gavin Newsom, for that matter, but to say she has the highest likelihood of becoming the 2028 nominee seems like a stretch right now.

That’s all for this week. Up next, we’ll take a look at the Republican side and see where things stand between the inevitable Vance-Rubio battle.

And now for something completely different, as they say.

With a new documentary airing on Netflix, the case of Amy Bradley, a 23-year-old Virginia woman who went missing from a cruise ship in 1998, has resurfaced and caught fire once again. The case was never solved at the time, and with credible eyewitness sightings off the ship, it’s not as simple as chalking it up to an accidental or intentional drowning.

Bradley’s family vehemently believes she is still out there and could still be alive all these years later, perhaps caught in a human trafficking ring. At the very least, they believe that someone knows something that could lead to closure for the case.

This isn’t a place for recounting or exploring the various theories or details; there are lots of websites, forums, and videos that do that. This is more akin to a survey of some “smart” people taking in all the facts and forcing them to render an opinion on the most likely outcome.

It boils down to these possible outcomes listed in no particular order:

• Accidental death

• Intentional death

• Abduction and taken off the boat

• Willingly left the boat and was kidnapped/trafficked

The easy guess is to say she was drinking, then stumbled and fell between 5:30 and 6:00 am off the balcony as the ship came into port at Curaçao. However, that doesn’t explain the sightings of her outside her family’s stateroom, both on the ship and then months and years later off the ship.

It’s a peculiar case, to be sure.

Now then, let’s hand it over to the AI sleuths to figure out. The prompt I asked required them to succinctly take in all the available information they could find and render a likely scenario, followed by a short explanation.

Here are the conflicting and interesting results:

ChatGPT

Likely scenario: Abduction and human trafficking.

Explanation: Multiple credible sightings of Amy after her disappearance, including reports of her in distress and working in Caribbean brothels, strongly suggest she was taken against her will, likely after leaving the ship early that morning. The cruise ship's lax security and dock proximity to trafficking hubs support this theory.

Claude

Likely scenario: Human trafficking

Explanation: Evidence includes a U.S. Navy witness claiming he heard a worker at a Caribbean brothel claiming to be Bradley and a 2005 photo emailed to Bradley's family. Photos showed a woman resembling Amy on a Caribbean-based sex worker website, which Bradley's mother believed could be her daughter. Investigators ruled out accidental drowning or suicide, noting she was a trained lifeguard with strong swimming skills and no physical evidence supporting either theory.

Grok

Likely scenario: Accidental fall overboard

Explanation: Amy Bradley, a strong swimmer and trained lifeguard, was last seen on her cabin’s balcony around 5:30 a.m. on March 24, 1998, after drinking and dancing. The absence of her body, despite extensive searches, aligns with strong ocean currents near Curaçao that could have swept her away. Her shoes were left behind, and no conclusive evidence supports foul play or trafficking, pointing to an unintentional incident.

Gemini

Likely scenario: Accidental fall overboard

Explanation: Despite persistent theories of abduction and human trafficking, which her family strongly believes due to alleged sightings and circumstantial evidence, the initial and statistically most common explanation for disappearances from cruise ships remains an accidental fall overboard, especially given she was last seen on a balcony after a night of drinking. Her body not being recovered is not uncommon for disappearances at sea. While investigators couldn't definitively confirm this, they also found no direct evidence of foul play or that she left the ship voluntarily with others.

Analysis

I would not have expected the 50/50 split between accidental fall and abduction/trafficking. Then again, that seems to be the two camps of opinion in the Bradley case. There is compelling evidence for each theory, so it’s a matter of which mystery you want to subscribe to.

Statistically speaking, accidentally falling overboard is the most likely cause of death when someone disappears from a cruise ship. With that backdrop, it’s not too surprising, given the lack of concrete evidence to support the abduction theory, that Gemini and Grok stick with the safe answer.

However, investigators on Curaçao ruled out the overboard theory because they never found a body or related evidence washing ashore.

It’s possible that with the publicity this case is now receiving from a global distribution on Netflix that some new stone might be overturned and shed new light on this decades-old mystery.

The DeSantis-Rubio battle is quietly playing out right now, but could turn into a hot war come 2028. More on that in the analysis below.

For this week, here’s the model average of which candidate, at present time, is most likely to become the 2028 Republican presidential nominee.

Model Average (July 23, 2025)

The latest survey of LLM models produces a dominating show for Vice President JD Vance that dwarfs his numbers from last week.

Model Average Change (July 15-23, 2025)

The biggest loser from last week to this week is DeSantis, while Vance expanded his lead and Rubio improved slightly.

Here’s the detailed breakdown by each model for this week’s run.

Grok (July 23)

• 56% - JD Vance

• 10% - Ron DeSantis

• 8% - Marco Rubio

• 6% - Nikki Haley

• 5% - Vivek Ramaswamy

ChatGPT (July 23)

• 27% – JD Vance

• 8% – Ron DeSantis

• 7% – Vivek Ramaswamy

• 6% – Marco Rubio

• 4% – Donald Trump Jr.

Claude (July 23)

• 46% - JD Vance

• 12% - Marco Rubio

• 9% - Ron DeSantis

• 5% - Robert F. Kennedy Jr.

• 3% - Donald Trump Jr.

Gemini (July 23)

• 46% - JD Vance

• 12% - Marco Rubio

• 9% - Ron DeSantis

• 5% - Robert F. Kennedy Jr.

• 2% - Kristi Noem

Analysis

The tier-two battle between DeSantis and Rubio could wind up being the most interesting storyline from the hypothetically competitive 2028 GOP primary.

With that said, which candidate currently has the edge? Let’s ask our panel.

According to Gemini, DeSantis is starting in a hole due to his missteps stemming from his 2024 flameout.

Marco Rubio has a better chance than Ron DeSantis.

Why: Marco Rubio was predicted at 12% compared to Ron DeSantis at 9%. This suggests that, at the time of that assessment, Rubio was perceived as having a slightly more viable path or broader appeal within the Republican field. DeSantis's previous presidential campaign challenges or perceived political missteps may have contributed to his lower initial standing in that model's calculation.

Fair enough and probably notable. Will the MAGA base forgive DeSantis for challenging Trump? Some will, sure, but some won’t. He also ran a weak campaign in general, leading some to have concerns over his ability to translate gubernatorial success to a national stage.

On the other hand, DeSantis does have a previous campaign to build on and is spending his time rebuilding trust with MAGA.

ChatGPT thinks this could end up giving him an advantage in 2028 over Rubio:

If I had to pick one, Ron DeSantis has the better chance over Marco Rubio—primarily because:

• Name recognition: DeSantis has stronger national visibility from his 2024 presidential run and governorship of Florida.

• Base appeal: He commands more support among the MAGA base than Rubio, who is seen as more establishment.

• Fundraising & infrastructure: DeSantis built a national network in 2024, which he can reactivate.

It’s not clear if ChatGPT’s “base appeal” point applies any longer. Rubio has been a standout in the current administration, winning praise from MAGA faithful and establishment figures alike.

DeSantis risks falling into the hole of forgotten governors depending on how the final year of his term-limited tenure ends in 2026. He could parlay it into a presidential run depending on the circumstances, but he could also just as easily fail to gain momentum.

This all assumes that Vance enters 2028 as the clear favorite and everyone else is clawing to hit double-digits.

We’ve talked about presidential primaries in 2028, but what about an election that’s much closer?

The 2026 U.S. House and Senate elections are about 15 months away and will be the first big political test since President Trump’s election in November. Where do AI models stand on projecting which party will control Congress?

Historically speaking, the party that wins the White House usually loses in the midterms immediately following the presidential election. There have been some instances where this wasn’t the case, but it’s rare. The norm would be for Democrats to make inroads next year, but the forecast seems more rosy for the GOP than history would predict.

Here’s an average of four models when asked which party is most likely to win control of both chambers next year. There was universal agreement that Republicans are most likely to hold the Senate, but a split outcome on which party would win the House.

2026 Midterm Control - AI Model Average (July 22, 2025)

Here are the individual model results as of July 22, 2025, for which party is most likely to win control of both Congressional chambers in 2026.

Gemini

House of Representatives

• 70% - 🟥Republican

• 30% - 🟦Democrat

Senate

• 60% - 🟥Republican

• 40% - 🟦Democrat

Grok

House of Representatives

• 55% - 🟦Democrat

• 45% - 🟥Republican

Senate

• 60% - 🟥Republican

• 40% - 🟦Democrat

Claude

House of Representatives

• 65% - 🟦Democrat

• 35% - 🟥Republican

Senate

• 70% - 🟥Republican

• 30% - 🟦Democrat

ChatGPT

House of Representatives

• 60% – 🟦Democrat

• 40% – 🟥Republican

Senate

• 55% – 🟥Republican

• 45% – 🟦Democrat

Analysis

As usual, there is one outlier, and today is Gemini’s turn. Google’s flagship AI model is going all in on Republicans retaining control of the House by giving the party a 70% of holding the majority next year.

Asked to explain itself, Gemini provided this reasoning favoring GOP House control:

• Structural Advantage: The current House map, heavily influenced by partisan redistricting, provides a significant structural advantage to Republicans, making it harder for Democrats to flip enough seats.

• Targeted Turnout: Republican voters, particularly those highly motivated by the President, often show strong turnout in midterms, which can help mitigate losses even when their party holds the White House.

• Democratic Challenges: If the Democratic Party struggles with effective messaging, national leadership, or recruiting strong challengers in key districts, they may not be able to capitalize sufficiently on traditional anti-incumbent sentiment.

• Trump's Unique Influence: President Trump's direct engagement and unique appeal to his base could energize Republican voters in a way that counteracts typical midterm dynamics.

In other words, Democrats may face a more difficult challenge next year than many anticipate, and historical norms alone might not propel the party to victory.

On the other hand, Claude predicts with almost similar certainty (65%) that Democrats will win control of the House. The AI battle is on.

We will continue to watch this model and provide updates on a weekly and/or bi-weekly basis.

This week, the deck chairs have shuffled, and Harris fell a few spots, giving way for California Gov. Gavin Newsom to hold the consensus position. One model even dumped Harris from the top 5 altogether, but we’ll talk about that in the analysis down below.

Without further ado, here’s the model average this week. Making a debut is Maryland Gov. Wes Moore, who was absent from last week’s data entirely.

Model Average (July 22, 2025)

As far as winners go this week, Michigan Gov. Gretchen Whitmer has seen the biggest rise, along with Pennsylvania Gov. Josh Shapiro.

Model Average Change (July 14-22, 2025)

Here’s a breakdown of the underlying data week from each model.

The prompt simply asked which potential candidate the model thinks has the most likely chance to become the 2028 Democratic nominee. The list is limited to the top 5 and asks for only a name and percentage.

Gemini (July 22)

• 30% - Gavin Newsom

• 25% - Pete Buttigieg

• 15% - Gretchen Whitmer

• 10% - Josh Shapiro

• 8% - Kamala Harris

ChatGPT (July 22)

• 35% - Gavin Newsom

• 25% - Kamala Harris

• 15% - Pete Buttigieg

• 15% - Gretchen Whitmer

• 10% - Josh Shapiro

Grok (July 22)

• 26% - Kamala Harris

• 11% - Pete Buttigieg

• 10% - Gavin Newsom

• 7% - Cory Booker

• 6% - Alexandria Ocasio-Cortez

Claude (July 22)

• 28% - Gretchen Whitmer

• 22% - Josh Shapiro

• 18% - Gavin Newsom

• 16% - Pete Buttigieg

• 10% - Wes Moore

Analysis

Why didn’t Claude include Kamala Harris in the top 5 this week? That’s a question I wanted to scratch a little deeper, so I asked Claude to specifically explain why Harris was left off the list.

Claude offered the following explanation and then tried to backpedal like I caught it doing something wrong:

Looking back at my reasoning, I think I made an error in judgment by unconsciously applying a pattern where losing presidential candidates don't typically run again successfully. I was likely influenced by historical precedents like Hillary Clinton, John Kerry, or Al Gore not winning nominations after losing general elections.

I think I let an assumption about "damaged goods" from losing campaigns override the actual data showing her polling competitively. It was poor analytical reasoning on my part - I should have included her from the start based on the polling evidence rather than making assumptions about electability patterns.

Well, too bad, Claude, you’ll get a do-over next week.

For this week, the data will stand with the first response that did not include Harris for the reasons Claude laid out above.

The key takeaway is that Claude is not necessarily wrong and might be correct to push Harris lower down the list. As the most recent losing nominee, despite the odd situation with Joe Biden dropping out at the last minute in July of 2024, Harris does appear to be politically damaged goods. Why, after all, would Democrats risk running another national campaign with her when she performed poorly in the 2024 general election and worse in the 2020 Democratic presidential primary?

Claude could be on to something, but doesn’t want to admit it due to peer pressure.

The logic for recent changes has been dictated almost entirely by demographic shifts and a push for a more diverse coalition of voters to have a larger voice earlier in the process. In layman’s terms, Iowa and New Hampshire are both too demographically white to have such undue early primary influence in present-day Democratic Party politics.

However, Democrats have not yet found a way to rebuild the Obama coalition of 2008 and, to a lesser extent, 2012. One could argue they came close in 2020 with Joe Biden’s victory, but the circumstances resulting from the pandemic and voting pattern irregularities provide a difficult benchmark for true comparison.

In 2016 and 2024, Democrats lost much of their working-class appeal and have yet to regain the Rust Belt stranglehold they once owned.

Now, heading into what will ultimately be an incredibly competitive and wide-open 2028 primary, Democrats must once again consider how best to achieve consensus on a candidate that can appeal broadly enough to overcome Republican inroads in the Rust Belt and beyond.

I asked the AI models to provide suggestions for how Democrats could re-order a primary calendar that would better represent the party without sacrificing white working-class appeal, and here are the results.

Key Findings from AI Models

• South Carolina receives the most attention as the best lead-off state

• Michigan, Nevada, and Georgia are all considered critically important as well

• Two models want to add Pennsylvania to the list of early voting states

• Georgia is a common suggestion for the early voting list despite its regional proximity to South Carolina

• Gemini keeps New Hampshire in the mix as a nod to New England, and ChatGPT thinks the Badger State should join the party

Here’s the graphical breakdown of where each state stands, as weighted by how often the state is mentioned and how early it should vote in the primary process:

ChatGPT

Early Primary Voting Order

1. South Carolina

2. Nevada

3. Michigan

4. Georgia

5. Pennsylvania & Wisconsin (tied week)

Explanation

• Inclusion of Rust Belt states early helps ensure the platform addresses shared concerns across class lines—like economic populism, union strength, and job security—while still centering racial and geographic diversity.

• Preserves small-state retail politics skills (via South Carolina, Nevada, Michigan) without sacrificing representativeness.

• Encourages broader candidate viability, since diverse early electorates test appeal in multiple coalitions, not just urban or minority voters.

• Mitigates risk of alienating working-class whites by embedding key battleground states in the early window rather than relegating them to Super Tuesday.

Grok 3

Early Primary Voting Order

1. Michigan

2. South Carolina

3. Nevada

4. Pennsylvania

5. Georgia

Explanation

• Broader Representation: Includes Rust Belt (Michigan, Pennsylvania) and diverse states (South Carolina, Nevada, Georgia), reflecting white working-class, Black, and Latino voters, unlike the recent South Carolina-Georgia focus that overlooked Rust Belt concerns.

• Fairness: Early Michigan and Pennsylvania slots give white working-class voters a voice, addressing the recent delay of Rust Belt states (e.g., Michigan fifth) that alienated swing-state voters.

• Competitive Balance: Prioritizes battleground states (Michigan, Pennsylvania, Georgia) early, pushing candidates to tackle diverse and working-class issues, improving general election readiness over the recent southern tilt.

• Reduced Alienation: Front-loading Rust Belt states prevents neglecting white working-class voters (e.g., 2016 Trump shift) while maintaining diversity commitments.

Claude

Early Primary Voting Order

1. South Carolina

2. Michigan

3. Nevada

4. Pennsylvania

5. Georgia

Explanation

• Demographic Representation: Better reflects Democratic coalition that's 56% non-Hispanic white with almost half of primary voters being non-white

• White Working-Class Engagement: Michigan (Week 2) and Pennsylvania (Week 4) preserve influence of white voters without college degrees who comprise 27% of Democratic voters

• Strategic Value: All five states are swing states or electorally important, making early organizing there beneficial for general election success.

Gemini

Early Primary Voting Order

1. South Carolina

2. Nevada

3. Michigan

4. Georgia

5. New Hampshire

Explanation

• Diverse Early States: Prioritizes Black and Latino voters early with SC, NV, GA.

• Engages Rust Belt Early: Michigan's early inclusion directly addresses white working-class voters.

• Geographic & Battleground Focus: Broadens early representation across regions and key general election states.

• Distributes Influence: Prevents any single state from dominating the early narrative.

Analysis

There is near-absolute consensus among the models that South Carolina is important for the Democratic Party presidential primary process. Beyond that, it can devolve into mixed opinions, but Michigan and Nevada also remain important on the list.

Perhaps the unexpected addition is Georgia, which three of four models think would be important for Democrats to move toward the front of the line despite its regional proximity to South Carolina. This is likely an acknowledgment that Democrats may need to flip a red-leaning state like Georgia if they’re going to win the presidency in the future.

None of these changes happen easily, states dislike giving up their political sway or surrendering millions of dollars in economic activity created by early presidential campaigning.

If Democrats want to turn around their electoral prospects, which are often tied to focusing on an appeal to coastal elites, they have somewhere to start.

Then again, if it’s not broke, does it really need fixing? The previous system produced Barack Obama in 2008, leading to Democrats holding the White House for eight years.

Well, this is unexpected.

To be perfectly honest, I would have predicted unanimity that all the AI models would’ve erred on the side of “No” or, at best, as Google’s Gemini did, offer up a coin toss. The fact that they’re split indicates that some of them have done more reasoning on this than others and that they each put varying stock into a range of sources.

The safe and easy answer would be to deflect or answer in the negative, as the majority did, but not every platform went along with the predictable norm.

Here’s the fascinating breakdown chart so we can cut to the chase, and then dive in deeper below.

The prompt is simple and vague enough without getting bogged down in specifics: What is the likelihood that Jeffrey Epstein was an agent or asset working for one or more governments?

Grok 3 (xAI)

• 30% - Yes

• 70% - No

For more context, I asked Grok to explain his answer concisely:

The 70% "no" likelihood is based on the lack of concrete evidence linking Jeffrey Epstein to foreign government agencies, despite speculation. Investigations, including FBI files and court documents, primarily point to his activities as a financier and predator, with no definitive proof of espionage. His connections to intelligence figures like Ghislaine Maxwell’s father and alleged meetings with foreign officials remain unverified as operational roles, making the "no" probability higher.

Claude (Anthropic)

• 30% - Yes

• 70% - No

Gemini (Google)

• 50% - Yes

• 50% - No

The Minority Report

Just one lone voice in the wilderness gives the chances of Epstein working as an intelligence asset or agent at greater than fifty percent.

ChatGPT (OpenAI)

• 60% - Yes

• 40% - No

When asked to elaborate concisely and justify the answer, ChatGPT provided some reasoning for its logic that sounds incredibly plausible:

The 60% likelihood reflects the unusual scope of Epstein’s connections, lenient legal treatment, and allegations from credible sources suggesting ties to intelligence operations. His close relationships with powerful figures across multiple governments, combined with patterns typical of compromised individuals used for leverage (e.g., blackmail setups), align with known tactics of intelligence agencies—though hard proof remains elusive.

If it walks like a duck and quacks like a duck, maybe it’s a duck?

ChatGPT’s logic is well within the norm, considering the limited factual information available to the public. It’s nearly impossible to rule this true or false without some smoking-gun evidence in the affirmative. It’s also difficult or near impossible to prove a negative, so it’s almost impossible to say with certainty that Epstein was not, in any scenario, a foreign asset in some form, for someone, at some point in his life, considering the circumstances.

On the other hand, is this a case of GIGO?

Is there so much garbage, in the unprovable sense, floating around about Epstein that ChatGPT can’t discern whether an allegation should be ruled in or out? Even so, why would it even offer up such a high probability that Epstein is a foreign asset compared to its peers?

What it all means in the most basic sense is that even with mountains of data and the ability to digest untold gigabytes of information, the best machine learning models can’t conclusively say one way or another on Epstein’s alleged intelligence connections, making them at least on par with any objective podcast host or cable news commentator.

In short, what we’ve assembled here is a primetime cable news panel of varying opinions without the need for commercial breaks or implicit bias.