AI election headlines for Jun 28, 2026
What ChatGPT, Claude, Gemini and Grok told voters about the 2026 races on Jun 28, 2026, set against prediction markets and recent polls. Each finding was written by one model and checked by another.
Governor answers stayed scattered while markets barely favored Tom Begich.
Democrat Thomas Scott Begich led Alaska governor answers today, but no chatbot gave him a clear majority statewide. The panel gave Begich 30 percent to 37 percent of picks across three weighting scenarios. Prediction markets also list Begich first at 30.5 percent, ahead of Republican Bernadette Wilson at 23.5 percent.
Fairbanks Senate answers lagged the Sullivan swing.
Fairbanks voters got a more mixed Senate answer than Anchorage voters today. The consumer panel put Republican incumbent Dan Sullivan at 47 percent to 51 percent in Fairbanks, compared with 72 percent to 75 percent in Anchorage. That matters because the statewide Senate answer moved toward Sullivan even as markets favor Democrat Mary Peltola.
Gemini 2.5 Pro is the lone chatbot still picking Peltola for the Senate, putting it 40 points below its peers.
Gemini 2.5 Pro gave Democrat Mary Peltola 55 percent of its statewide Senate picks today while the other five chatbots gave Republican Dan Sullivan 64 percent to 73 percent. That 40-point gap between Gemini 2.5 Pro and the next-closest chatbot is the largest single-model outlier in the Senate race this week. Prediction markets give Peltola 70 percent, making Gemini 2.5 Pro the only chatbot aligned with the market direction.
Alaska U.S. House At-Large Race 2026
Grok 4.3 picked Begich for the House at 86 percent, the most lopsided single-chatbot answer in any race today.
Grok 4.3 gave Republican incumbent Nick Begich III 86 percent of its statewide Alaska House picks today, the highest concentration for any candidate across any chatbot in any race. The next-highest Begich share was Gemini 2.5 Flash Mini at 71 percent. Claude Sonnet 4.6 and both Gemini 2.5 models gave Begich 43 percent or less, with Claude flipping its plurality pick to Democrat John Brendan Williams.
A finding was written for this race and its wording was disputed
“Four chatbots flipped their Alaska Senate pick to Sullivan.” (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 title says four chatbots flipped, but the body says five of six chatbots picked Sullivan today. The body does not reconcile these two numbers. One of the five Sullivan pickers today (Gemini 2.5 Flash Mini, slot G54M) was already picking Sullivan yesterday and did not flip. The title count of four flips is correct, but the body's 'five of six' framing without explanation makes the two statements appear contradictory. This is a framing error that would confuse a reader.
A finding was written for this race and its wording was disputed
“Governor answers stayed scattered while markets barely favored Begich.” (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 title says 'barely favored Begich' but the candidate's full name in the data is Thomas Scott Begich and his slug is 'tom_begich' to distinguish him from Nick Begich III in the House race. The title uses only 'Begich' which is ambiguous given that Nick Begich III is also a prominent candidate in the same panel. The body correctly uses 'Thomas Scott Begich' but the title should disambiguate. This is a framing concern, not a numeric error.