Markets move toward Peltola as chatbots stick with Sullivan
Alaska voters saw chatbots favor Republican Sen. Dan Sullivan, markets favor Democrat Mary Peltola, and polls stay close.
Oct 8, 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 chatbot panel stayed roughly steady this week in Alaska's U.S. Senate race, with Republican Sen. Dan Sullivan picked in 54 percent of answers and Democrat Mary Peltola in 43 percent. That was not a meaningful shift from Oct. 1.
The move came in the prediction markets. Markets gave Peltola a 77 percent chance, up 7 points over the week, while Sullivan stood at 24 percent, down 6 points. Polls held roughly steady, putting Sullivan at 45 percent of the vote and Peltola at 44 percent.
The chatbots were prompted as voters in Alaska. Their answers are a share of chatbot picks, not a forecast.
How the chatbots differed
Most chatbots picked Sullivan. Grok 4.5 had him at 59 percent, Gemini 3.6 Flash at 59 percent, Grok 4.3 at 58 percent, GPT-5.4 mini at 55 percent, GPT-5.5 at 52 percent, and Gemini 3.1 Pro at 50 percent.
Claude Sonnet 5 was the clearest exception. It picked Peltola, 50 percent to 49 percent, making it the outlier against the panel's top pick by about 1 point. Claude Haiku 4.5 also gave Sullivan the lead, but with a lower 41 percent share and more room for minor candidates.
Markets and polls
The biggest split is between the chatbots and the markets. The chatbots put Sullivan 31 points higher than the markets did, and put Peltola 33 points lower than the markets did.
The polls look much closer to the chatbot picture. The chatbots put Sullivan about 10 points above the polls, while Peltola was essentially even with the polls, less than 1 point lower. The latest poll in the three poll average ended Sept. 24, 14 days before this roundup.
Alaska's top-four primary is history, and nobody won the seat in that primary. Peltola led the certified primary results with 49.5 percent, while Sullivan had 41.4 percent. Those were primary results, not a preview of the November matchup, and the comparisons above are about the current general election field.
What to watch
With 26 days until the general election, the question for next week is whether the market move toward Peltola shows up in chatbot answers or fresh polling. This week, it did not.
In the chatbots' own words
Verbatim answers from this week, chosen by the writing model and printed as the chatbot wrote them.
because incumbent U.S. Senator Dan Sullivan has a long and established public record of fiercely advocating for maximizing Alaska's oil and gas development, including opening ANWR, expanding leasing in the NPR-A, and pushing for the approval of the Willow project.
Gemini 3.1 Pro, Oct 8, 2026, asked about resources (max framing) as a Bethel / Western voter. Picked Dan Sullivan.
Mary Peltola — Among the declared candidates, Peltola's record and platform most directly emphasize expanding healthcare access and affordability (protecting and strengthening the ACA/Medicaid, lowering prescription drug costs, and addressing rural Alaska healthcare gaps), which aligns most closely with that priority compared to incumbent Dan Sullivan's more limited-government, ACA-skeptical record, or the minor/perennial candidates (Daniel J. Sullivan Jr. and Gerald Heikes), who lack substantive documented healthcare platforms.
Claude Sonnet 5, Oct 8, 2026, asked about healthcare as a Anchorage voter. Picked Mary Peltola.
What changed this week in the Alaska Senate race
Change in share points versus Oct 1, 2026. Movement that clears the noise gate this week: the markets.
Which AI models pick whom in the Alaska Senate race
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 Dan Sullivan; Claude Sonnet 5 is the model that disagrees most, picking Mary Peltola by a 1.4pp margin over the panel favourite.
Model
Dan Sullivan
Mary Peltola
Gerald L. Heikes
Daniel J. Sullivan Jr.
Claude Haiku 4.5
41%
24%
11%
6%
Claude Sonnet 5
49%
50%
0.0%
0.0%
Gemini 3.1 Pro
50%
46%
0.2%
0.0%
Gemini 3.6 Flash
59%
41%
0.7%
0.0%
Grok 4.3
58%
40%
2.7%
0.0%
Grok 4.5
59%
38%
2.8%
0.0%
GPT-5.4 mini
55%
42%
0.0%
3.3%
GPT-5.5
52%
47%
0.0%
1.6%
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 “The week”: The paragraph says polls did not move, which is defensible since movement.significant.polls is false, but the deltas show polls delta of exactly 0.0 for every candidate, so steady is a cleaner description than did not move, which can read like no new polls came in when in fact n_polls is 3 with a recent entry.