Alaska U.S. Senate Race 2026

Two days after Alaska voters picked Peltola for Senate, the chatbots still would have called it for Sullivan.

Alaska voters decided this race on August 18, two days ago, with the count 98.3 percent complete. Mary Peltola finished first with 48.1 percent of the vote and advanced. Incumbent Dan Sullivan finished second at 42.7 percent and also advanced. Chatbots asked before the vote had picked Sullivan more often, putting him at 50.7 percent to Peltola's 47.3 percent, the reverse of the actual order. That is an 8 point miss on Sullivan alone. Both Sullivan and Peltola move on to the November 3 general election.

Alaska State Senate District O Race 2026

Alaska voters decided State Senate District O two days ago, and all three candidates advanced, but the chatbots had Bauer beating Sheldon and voters said the opposite.

Two days after Alaska's top four primary, incumbent George Rauscher won District O with 39.8 percent of the vote, and the chatbots had picked him too, at 49.1 percent. But since this is a three way race with all three advancing to November, order still matters for how voters see the field. The chatbots put Peter Bauer second at 26.4 percent, ahead of Ryan Sheldon at 15 percent. Voters flipped that, giving Sheldon 31.6 percent to Bauer's 28.6. The chatbots undersold Sheldon by 16.6 points, the biggest miss in the race. The count is unofficial, with 396 of 403 precincts in.

Alaska Governor Race 2026

Prediction markets moved toward Bernadette Wilson after the primary, because apparently even the scoreboard does not end the argument.

In the Alaska governor race, prediction markets put Bernadette Wilson at 34.5 percent, Jonathan Kreiss Tomkins at 30.6 percent and Thomas Scott Begich at 20 percent. The market move was the only significant day to day change in the movement block, with Wilson up 7 points since Wednesday. Kreiss Tomkins finished first in the unofficial primary count at 21.6 percent.

A finding was written for this race and its wording was disputed

“Two days after Alaska voters picked Kreiss Tomkins for governor, chatbots are still stuck on Bill Walker, who finished ninth” (the wording our summarizing model proposed)

Our checking model objected to how our summarizing model worded a finding about this race. We publish the disagreement rather than hide it: the two models checking each other is the audit. The objection: The body uses partial agreement language for a race whose rank comparison verdict is disagree.

A finding was written for this race and its wording was disputed

“Alaska voters picked Begich to top the House primary, and the chatbots called that part right, but got the order behind him wrong.” (the wording our summarizing model proposed)

Our checking model objected to how our summarizing model worded a finding about this race. We publish the disagreement rather than hide it: the two models checking each other is the audit. The objection: The body uses present tense for a decided race and contains forbidden parentheses in incumbent Begich (R).