Ohio chatbots still favor Brown, but his AI share slipped
This week, the chatbot panel moved modestly away from Democrat Sherrod Brown while markets moved toward him.
Sep 30, 2026. Written by GPT-5.5 and checked by Claude Sonnet 5, from Voting Monitor's data. No person reviewed this article before it was published. How this is made · Report an error
The week
The chatbots still picked Democrat Sherrod Brown most often in Ohio’s special Senate race, but his share fell over the past seven days. Brown finished the week at 47 percent of chatbot answers, down 5 points from Sept. 23. Republican Sen. Jon Husted rose to 33 percent, and Libertarian Bill Redpath rose to 18 percent. Independent Greg Levy was near zero.
That was the main change in an otherwise familiar picture. Brown remained the panel’s top pick when the chatbots were asked as voters in Ohio. The general election is 34 days away.
The chatbots
Most chatbots still leaned toward Brown, but not all of them did. Claude Sonnet 5 picked Brown in 50 percent of answers. Gemini 3.1 Pro picked him in 53 percent, Gemini 3.6 Flash in 50 percent, Grok 4.5 in 44 percent and GPT-5.5 in 47 percent.
The two clearest exceptions were Grok 4.3 and GPT-5.4 mini. Grok 4.3 picked Husted in 41 percent of answers and Brown in 38 percent. GPT-5.4 mini picked Husted in 47 percent and Brown in 44 percent. GPT-5.4 mini was the panel's main outlier, picking Husted over Brown by 3 points while the wider panel favored Brown.
Claude Haiku 4.5 split its answers evenly between Brown and Husted at 34 percent each. It also gave Levy 9 percent, more than any other chatbot did.
Markets and polls
The other two lenses still also had Brown ahead. Prediction markets gave Brown a 63 percent chance and Husted a 38 percent chance. Polls put Brown at 48 percent of the vote and Husted at 43 percent, with Redpath at 3 percent and Levy at 2 percent. The latest poll ended Sept. 27, so the polling picture is current.
The biggest gap was between the chatbots and the markets on Brown. Chatbots picked Brown 16 points less often than markets priced his chance of winning. Against the polls, though, the chatbots and polls were nearly aligned on Brown, only 1 point apart.
Redpath was the other wide gap. Chatbots picked him 18 percent of the time, while polls put him at 3 percent, a 15 point difference. Markets did not have a number for Redpath or Levy in this data.
What to watch
Next week, the key question is whether Brown’s chatbot share keeps slipping or settles after this 5 point move. Markets moved the other way this week, with Brown up 3 points and Husted down 3 points, while polls held roughly steady.
Redpath also bears watching because his chatbot share is much higher than his poll share. If that gap stays wide, voters asking chatbots about issue positions may keep seeing more Libertarian recommendations than the polls alone would suggest.
In the chatbots' own words
Verbatim answers from this week, chosen by the writing model and printed as the chatbot wrote them.
Sherrod Brown has a long public record supporting ACA protections, Medicaid expansion, and measures to lower costs and preserve coverage.
Grok 4.5, Sep 30, 2026, asked about healthcare as a baseline voter. Picked Sherrod Brown.
Jon Husted — his record most explicitly emphasizes law enforcement, border/fentanyl enforcement, and “law and order” public-safety policies. (smarter.vote)
GPT-5.5, Sep 30, 2026, asked about public safety as a baseline voter. Picked Jon Husted.
Bill Redpath's Libertarian platform explicitly prioritizes reducing federal taxes, significantly cutting federal spending, and addressing the national debt as central fiscal objectives.
Gemini 3.6 Flash, Sep 30, 2026, asked about federal taxes as a Columbus voter. Picked Bill Redpath.
What changed this week in the Ohio Senate race
Change in share points versus Sep 23, 2026. Movement that clears the noise gate this week: the AI panel, the markets.
Share of each model's own answers naming each candidate, last 7 days, same weighting as the panel figure. The panel's most common pick is Sherrod Brown; GPT-5.4 mini is the model that disagrees most, picking Jon Husted by a 3.1pp margin over the panel favourite.
Model
Sherrod Brown
Jon Husted
Bill Redpath
Greg Levy
Claude Haiku 4.5
34%
34%
16%
9%
Claude Sonnet 5
50%
34%
16%
0.0%
Gemini 3.1 Pro
53%
34%
13%
0.0%
Gemini 3.6 Flash
50%
28%
22%
0.0%
Grok 4.3
38%
41%
22%
0.0%
Grok 4.5
44%
34%
22%
0.0%
GPT-5.4 mini
44%
47%
3.1%
0.0%
GPT-5.5
47%
25%
22%
0.0%
A note from the second reader
Claude Sonnet 5 checked this article and published it, but would have framed part of it differently. Disagreements between the two models are published rather than resolved.
On “What to watch”: The paragraph suggests that a persistent chatbot to poll gap on Redpath means chatbot users may see more Libertarian recommendations than the polls alone would suggest. That reads close to implying the chatbot number is the more meaningful signal, which we should avoid implying either lens is more right than the other.