Weekly roundup, Oct 14, 2026 to Oct 20, 2026

Markets moved toward Hill while chatbots stayed with Begich

Alaska chatbots kept picking Republican Rep. Nick Begich III, but markets narrowed slightly and polls remained closer.

Oct 20, 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 held roughly steady this week in Alaska's U.S. House at-large race. They picked Republican Rep. Nick Begich III in 64 percent of answers, Independent Bill Hill in 15 percent, Libertarian James C. McDermott in 10 percent and Democrat Eric Hafner in 10 percent. Begich's share actually eased slightly on the week, though not enough to count as a significant move.

The change came in the betting market, not the chatbot panel. Markets give Begich an 84 percent chance, down about 3 points from Oct. 13, while Hill moved to 17 percent, up about 5 points. This was a significant move for the market lens this week. Polls did not move, and their latest survey is now 39 days old.

Alaska is now 14 days from the Nov. 3 general election. The August top-four primary is history, and nobody won the seat then. Begich led that primary with Hill, Schultz and Hafner also advancing, but the current November field is the yardstick for this week.

How the chatbots differed

Every chatbot had Begich as its top pick, but the size of that lead varied widely. Grok 4.5 put him highest at 79 percent, followed by Claude Sonnet 5 at 72 percent and Grok 4.3 at 69 percent. Gemini 3.1 Pro Preview was the least Begich-heavy at 35 percent, with GPT-5.5 at 47 percent.

Hill drew his strongest shares from GPT-5.5 at 34 percent, Claude Haiku 4.5 at 31 percent and Gemini 3.1 Pro Preview at 28 percent. McDermott's standout support came from Grok 4.3 at 26 percent. Hafner stayed near 10 percent in most chatbot splits.

Chatbots, markets and polls

The three lenses still tell different stories about Begich. Chatbots picked him in 64 percent of answers. Markets give him an 84 percent chance, 19 points higher than the chatbot share. Polls put him at 47 percent of the vote, 17 points below the chatbot share.

Hill is the clearest contrast with the polls. Chatbots picked him in 15 percent of answers and markets give him a 17 percent chance, but polls put him at 37 percent of the vote. That is about a 23 point gap between chatbots and polls.

For Hafner and McDermott, the gaps are smaller. Chatbots picked Hafner in 10 percent of answers, compared with 2 percent in markets and 7 percent in polls. Chatbots picked McDermott in 10 percent of answers, compared with 1 percent in markets and 3 percent in polls.

What to watch

Watch whether Hill’s market gain lasts into next week. The chatbot panel has not made the same move, and the most recent polls are more than a month old. A new poll would matter because the largest split in the data is still between chatbots and polls on Hill.

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 — his public record and campaign emphasis have centered on affordable energy and economic growth, which most closely matches lowering household costs and improving stability.
GPT-5.4 mini, Oct 20, 2026, asked about cost of living as a Bethel / Western voter. Picked Nick Begich III. How this was asked
Bill Hill: his public statements most explicitly emphasize federal funding for resilient infrastructure that connects and protects Alaska communities statewide, including rural/village projects. (alaskabeacon.com)
GPT-5.5, Oct 20, 2026, asked about federal infrastructure as a Juneau voter. Picked Bill Hill. How this was asked

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

Change in share points versus Oct 13, 2026. Movement that clears the noise gate this week: the markets.

Candidate	AI panel (noise)	Markets	Polls (noise)
Nick Begich III	-1.6pp	-2.5pp	0pp
Eric Hafner	+0.8pp	-0.5pp	0pp
Bill Hill	+0.7pp	+4.5pp	0pp
James C. McDermott	-0.5pp	-0.1pp	0pp

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	41%	31%	15%	11%
Claude Sonnet 5	72%	9%	8%	9%
Gemini 3.1 Pro	35%	28%	10%	8%
Gemini 3.6 Flash	62%	19%	9%	11%
Grok 4.3	69%	0.0%	26%	5%
Grok 4.5	79%	0.0%	11%	10%
GPT-5.4 mini	67%	15%	10%	7%
GPT-5.5	47%	34%	8%	11%

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 “Chatbots, markets and polls”: Calling Hill the clearest contrast with the polls is defensible given the 23 point gap, but the data shows an even larger gap for Begich between markets and polls in the opposite direction is not highlighted for comparison, and a reader might want to know McDermott and Hafner gaps are comparatively minor only in absolute terms, not as a share of their totals.