Skip to main content
Voting Monitor2026 Senate toss-ups

Weekly roundup, Sep 30, 2026 to Oct 6, 2026

Chatbots stay with Begich as markets price him higher

This week, AI answers, betting markets and polls continued to tell different stories about Alaska's House race.

Oct 6, 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

A steady week

The chatbots still picked Republican incumbent Nick Begich III most often in Alaska's at-large House race. His share of chatbot answers was 62 percent, compared with 18 percent for Independent Bill Hill, 10 percent for Libertarian James C. McDermott and 7 percent for Democrat Eric Hafner.

The week did not bring a clear move in any lens. Compared with Sept. 29, the chatbot panel, the betting markets and the polls all held roughly steady. That is the main difference from last week's roundup, when McDermott had gained ground in AI answers.

The race is now in the general election phase, 28 days before Nov. 3. The August top-four primary is history. Begich led that primary with 44.6 percent, and Hill followed with 32.5 percent, but nobody won the seat in that primary.

The chatbots differed on the margin

Every chatbot in the panel put Begich first when prompted as a voter in Alaska, but the size of that lead varied. GPT-5.4 mini put Begich at 78 percent, Grok 4.5 at 77 percent and Grok 4.3 at 66 percent. GPT-5.5 was the least Begich-centered, putting him at 42 percent and Hill at 36 percent.

GPT-5.5 gave Hill his strongest share at 36 percent. Grok 4.3 gave McDermott his strongest share at 23 percent. Hafner's best showing was 12 percent, also with GPT-5.5. No single chatbot was marked as the panel's outlier this week.

Markets are higher on Begich than AI

The betting markets gave Begich an 87 percent chance, 24 points higher than the chatbots' 62 percent share of answers. Markets gave Hill a 13 percent chance, close to his 18 percent chatbot share.

Polls told a closer story. The poll average put Begich at 47 percent, Hill at 37 percent, Hafner at 7 percent and McDermott at 3 percent. The latest poll ended Sept. 11, so the poll average is useful standing context rather than a fresh break.

The widest poll gaps were for the two leaders. Chatbots put Begich 15 points above the polls and Hill 19 points below the polls. For Hafner, the chatbots and polls were nearly identical.

What to watch

Next week, watch whether any new poll narrows or confirms the gap between Hill's 37 percent poll share and his 18 percent chatbot share. Also watch whether the live market keeps Begich near 87 percent while the chatbot panel stays closer to 62 percent.

In the chatbots' own words

Verbatim answers from this week, chosen by the writing model and printed as the chatbot wrote them.

What changed this week in the Alaska U.S. House At-Large race

Change in share points versus Sep 29, 2026. No lens moved more than its noise gate this week; the changes below are within normal day-to-day variation.

CandidateAI panel (noise)Markets (noise)Polls (noise)
James C. McDermott-2.3pp+0.5pp0pp
Nick Begich III-1.3pp+2.0pp0pp
Eric Hafner+1.2pp-1.1pp0pp
Bill Hill+1.1pp-0.5pp0pp

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.

ModelNick Begich IIIBill HillJames C. McDermottEric Hafner
Claude Haiku 4.553%11%9%7%
Claude Sonnet 560%20%4.0%7%
Gemini 3.1 Pro52%29%10%7%
Gemini 3.6 Flash62%31%6%1.8%
Grok 4.366%0.2%23%11%
Grok 4.577%0.0%16%7%
GPT-5.4 mini78%4.2%10%8%
GPT-5.542%36%11%12%

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.