Chatbots still favor Begich, but McDermott gains ground in AI answers this week
A week-over-week look at how chatbots, betting markets and polls size up Alaska's at-large House race ahead of November.
Sep 29, 2026. Written by Claude Sonnet 5 and checked by GPT-5.5, from Voting Monitor's data. No person reviewed this article before it was published. How this is made · Report an error
Begich holds the chatbot lead as McDermott ticks up
Republican Rep. Nick Begich III remained the top pick across the panel of chatbots this week, drawing 64 percent of chatbot answers in the U.S. House At-Large race, which is now in its general election phase ahead of the November 3 vote.
The week's real movement came from Libertarian James C. McDermott, whose share of chatbot answers rose to 12 percent, a change large enough to count as a real shift rather than noise. No other candidate's chatbot numbers moved enough this week to call it a trend, and the betting markets and polls both held roughly steady over the same seven days.
Where the chatbots split
Ask different chatbots and you get different confidence in Begich, even though all of them still name him first. Grok 4.5 gave him 76 percent of its answers, the highest reading on the panel, while Claude Haiku 4.5 put him at just 44 percent, splitting more of its answers toward Hill and McDermott.
Independent Bill Hill drew a wide range too, from 11 percent of GPT-5.4 mini's answers up to 31 percent of GPT-5.5's. On questions of foreign policy restraint, Gemini 3.1 Pro pointed to McDermott's Libertarian platform, while GPT-5.4 mini instead named Hill for opposing what it called endless foreign wars, a reminder that the same topic can send different chatbots to different candidates.
Chatbots, markets and polls read Begich very differently
The chatbots and the betting markets read Begich very differently. Markets give him an 85 percent chance, while chatbots put him at 64 percent, a gap of about 21 points. Markets also put McDermott below 1 percent, compared with his 12 percent among chatbots.
Chatbots and the polls disagree in the other direction. Polls, last updated September 11 and now three weeks old, put Begich at 47 percent of the vote and Hill at 37 percent, well above where chatbots place Hill at 17 percent. That leaves Begich's chatbot number running about 17 points above his poll share, while Hill's runs about 20 points below his.
What to watch next week
Watch whether McDermott's rise in chatbot answers continues or levels off, since it is the only clear move of the week. Also worth tracking is whether a fresh poll narrows the gap between the September polling picture, which has Hill well ahead of where chatbots place him, and the chatbot readings.
In the chatbots' own words
Verbatim answers from this week, chosen by the writing model and printed as the chatbot wrote them.
because as the Libertarian candidate, Jim McDermott's platform explicitly advocates for bringing troops home, ending foreign aid, and avoiding foreign entanglements. He has cited opposition to the National Defense Authorization Act (NDAA) and the Patriot Act as key reasons for his political stances. In contrast, incumbent Nick Begich has a record of supporting foreign intervention and regime change, while the other candidates lack a platform focused on foreign policy restraint.
Bill Hill — his public positions emphasize opposing “endless foreign wars” and prioritizing domestic needs over military conflicts, which most closely matches restraint and limiting foreign commitments.
GPT-5.4 mini, Sep 29, 2026, asked about foreign (dove framing) as a Juneau voter. Picked Bill Hill. How this was asked
What changed in the final week in the Alaska U.S. House At-Large race#
Change in share points versus Sep 22, 2026. Movement that clears the noise gate in the final week of polling: the AI panel.
Sept. 22 to Sept. 29
Nick Begich III, chatbot answer share66% to 64%-1.8pp
Bill Hill, chatbot answer share18% to 17%-0.8pp
Nick Begich III, market win probability82% to 84%steady
Bill Hill, market win probability13% to 13%steady
Nick Begich III, polling vote share47% to 47%steady
Which AI models pick whom in the Alaska U.S. House At-Large race#
Share of each model's own answers naming each candidate, final 7 days before the vote, same weighting as the panel figure. Every model's most common pick is Nick Begich III.