2025 Virginia Governor's Race: RCP Polls, AI Predictions, and Reality
In this case, the RealClearPolitics average and AI models tend to agree it's Abigail Spanberger's race to lose, but there's still a chance for Winsome Sears to win
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.
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.
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
Seriously, this Roanoke College comparison is a little spooky:
Aug 2021: McAuliffe 46% – Youngkin 38% (+8 D)
Aug 2025: Spanberger 46% – Sears 39% (+7 D)