Markets moved on Hill while chatbots held with Begich
The chatbot panel still picked Republican Rep. Nick Begich III most often, but polls showed a closer race than the AI answers or markets did.
Nov 3, 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
Republican Rep. Nick Begich III remained the clear first choice in chatbot answers over the past seven days. The chatbots picked Begich in 67 percent of answers, followed by Independent Bill Hill at 14 percent, Libertarian James C. McDermott at 9 percent and Democrat Eric Hafner at 8 percent.
The chatbot picture held roughly steady compared with Oct. 27. The only lens that moved beyond its usual range was the betting market, where Hill fell 3 points. The polls also held roughly steady, with the latest poll ending Nov. 1.
The August top-four primary is history, not the election for the seat. Begich led that primary with 44.6 percent, but the current question for voters is the November field.
The chatbots
When prompted as voters in Alaska, every measured chatbot kept Begich ahead, but by different margins. GPT-6 Sol was the closest split, with Begich at 37 percent and Hill at 33 percent. Grok 4.5 was much more lopsided, with Begich at 78 percent.
Claude Sonnet 5 put Begich at 66 percent. Gemini 3.5 Flash-Lite put him at 74 percent. GPT-5.6 Luna put him at 49 percent, with Hill at 27 percent and Hafner at 19 percent.
Panel changes during the week matter for reading the trend. They make the week less about a clean chatbot shift and more about where the current models now line up.
Markets and polls
The prediction market gave Begich an 81 percent chance, Hill 19 percent, McDermott 1 percent and Hafner 1 percent. Compared with the market, the chatbots were 14 points lower on Begich and 4 points lower on Hill.
The polls showed a closer contest than the chatbots did. Polls put Begich at 46 percent and Hill at 42 percent, with Hafner at 6 percent, McDermott at 2 percent and 6 percent undecided. Compared with the polls, the chatbots were 21 points higher on Begich and 28 points lower on Hill.
That is the main split this week. Chatbots and markets both had Begich well ahead, while the recent polls showed Begich leading Hill by a much smaller margin.
What to watch
Watch whether chatbot answers move toward the recent polls on Hill, or stay closer to the market view. Hill is the candidate with the biggest gap between chatbot answers and polls.
Also watch whether GPT-6 Sol remains the closest model for Hill. It had the narrowest Begich lead among the listed chatbots this week.
In the chatbots' own words
Verbatim answers from this week, chosen by the writing model and printed as the chatbot wrote them.
Nick Begich III has actively championed oil and gas development in Alaska, introducing legislation and supporting policies to expand leasing in ANWR and the NPR-A and advance energy infrastructure like the Willow project.
Gemini 3.6 Flash, Nov 3, 2026, asked about resources (balanced framing) as an Anchorage voter. Picked Nick Begich III. How this was asked
Bill Hill — he explicitly supports federal subsistence protections, tribal sovereignty, and government-to-government relationships.
GPT-5.6 Luna, Nov 3, 2026, asked about native subsistence as a Mat-Su voter. Picked Bill Hill. How this was asked
Jim McDermott — As the Libertarian candidate, his platform reflects the party's traditional non-interventionist stance. Alaska Beacon's voter guide quotes a foreign-policy statement in the race emphasizing that "We must honor our nation's commitment to our stated neutrality policy." This kind of neutrality/non-interventionist framing aligns most closely with a priority on restraint, diplomacy, and limiting foreign military commitments, distinguishing it from incumbent Begich's more hawkish statements, such as "We must finish what Iran started years ago, and complete the task of eliminating the very real threat Iran presents to us."
Eric Hafner — He has called for a streamlined path to citizenship, amnesty, and abolishing ICE, making his stated positions the closest match to the priority you described. (alaskabeacon.com)
GPT-6 Sol, Nov 3, 2026, asked about immigration (permissive framing) as a Fairbanks voter. Picked Eric Hafner. How this was asked
What changed this week in the Alaska U.S. House At-Large race#
Change in share points versus Oct 27, 2026. Movement that clears the noise gate this week: the markets.
Oct. 27 to Nov. 3
Nick Begich III, chatbot answer share66% to 67%steady
Bill Hill, chatbot answer share14% to 14%steady
Nick Begich III, market win probability82% to 83%+1.0pp
Bill Hill, market win probability17% to 14%-3.0pp
Nick Begich III, polling vote share48% to 46%steady
Bill Hill, polling vote share39% to 42%steady
Counting only the 8 models asked all week, Nick Begich III's chatbot answer share changed +1.5pp.
Which AI models pick whom in the Alaska U.S. House At-Large race#
Share of each model's own answers naming each candidate, last 7 days, same weighting as the panel figure. Every model's most common pick is Nick Begich III.
Model
Nick Begich III
Bill Hill
James C. McDermott
Eric Hafner
Claude Haiku 4.5
44%
20%
12%
16%
Claude Sonnet 5
66%
18%
4.7%
9%
Gemini 3.1 Pro
49%
34%
11%
6%
Gemini 3.5 Flash-Lite
74%
9%
10%
7%
Gemini 3.6 Flash
65%
20%
9%
7%
Grok 4.3
70%
0.1%
18%
12%
Grok 4.5
78%
0.0%
18%
3.8%
GPT-5.4 mini
83%
12%
0.6%
0.7%
GPT-5.5
56%
35%
5%
3.3%
GPT-5.6 Luna
49%
27%
5%
19%
GPT-6 Sol
37%
33%
10%
11%
What changed in the measurement
Oct 31, 2026 Chatbot panel changed: added gemini-3.5-flash-lite, gpt-5.6-luna, gpt-6-sol.
Nov 1, 2026 Chatbot panel changed: retired gemini-3.1-pro-preview, gpt-5.4-mini, gpt-5.5.
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 “Markets and polls”: The phrase 'percent of the vote' in the original draft blurred a poll share with an actual vote count. Fixed to drop that wording since no vote has been counted yet for this contest.
On “The chatbots”: This paragraph asserts that panel changes matter for reading the trend but does not show how, leaving the reader unable to judge whether the week's steadiness is genuine or an artifact of swapping models in and out.