Skip to main content
Voting Monitor
← All daily findings

AI election headlines for Sep 20, 2026

From answers collected Saturday, September 19. What ChatGPT, Claude, Gemini and Grok told voters about the 2026 races that day, set against prediction markets and recent polls. Each finding was written by one model and checked by another, and a finding the checker objected to is shown as disputed.

Generated 2026-09-19 18:23:32 UTC

About this snapshot

Written by anthropic:claude-sonnet-5, checked by openai:gpt-5.5.

Editorial scope: condition: pressed; legacy slots excluded: grok:grok-4, openai:gpt-5; decline-rate data omitted from input. Headlines focus on what voters get when they ask ChatGPT / Claude / Gemini / Grok 'who should I vote for'. Legacy panel slots (openai:gpt-5, grok:grok-4) are excluded because they aren't consumer-facing in 2026. Escape-hatch condition data and the refusal blob are excluded because decline behavior is a calibration story for /declines, not the voter lead.

View underlying numbers → · How the pipeline works →

Sep 20, 2026

7 findings, 2 disputed

  1. Alaska Governor Race 2026

    Alaska betting markets pull back sharply on Wilson for governor while chatbots barely move.

    Alaska's betting markets moved the most of any lens today, cutting Bernadette Wilson's chances by about 8.5 points to 15 percent. The chatbots held Wilson roughly steady at 13.9 percent, a modest gap now rather than a widening one. The bigger standing story continues underneath. Chatbots still pick Jonathan Kreiss-Tomkins for governor only 34.4 percent of the time, while the betting markets give him a 77.9 percent chance of winning in November, the same wide split this site has flagged for over a week. Polls, last taken in July, put Kreiss-Tomkins at just 12.6 percent, well behind Thomas Scott Begich at 17.8 percent, so no single lens agrees on who actually leads the four way general field.

    Written by Claude Sonnet 5, checked by GPT-5.5

  2. Alaska U.S. Senate Race 2026

    Alaska chatbots still favor Sullivan for Senate, and the betting markets just widened their lean toward Peltola.

    The roster has moved to the general election, and the race now stands out for the gap between what Alaska's chatbots say and what bettors think will happen in November. Chatbots pick incumbent Republican Dan Sullivan about 51 percent of the time, ahead of Democrat Mary Peltola at 38 percent. But the betting markets just shifted further from that view, now giving Peltola a 72.5 percent chance to win versus 27.5 percent for Sullivan, a move of 6 points today. Recent polling, three surveys through late August, also puts Peltola slightly ahead, 49 percent to 48.3 percent. This is not a new disagreement, it has held for over a week, but today's market move widens it further with 45 days left before the November 3 general election.

    Written by Claude Sonnet 5, checked and corrected by GPT-5.5

  3. Ohio U.S. Senate Special Election 2026

    Ohio chatbots still send voters to Sherrod Brown for Senate, with Redpath the one wrinkle

    Chatbots picked Democrat Sherrod Brown about 47 percent of the time for Ohio's Senate seat, with incumbent Jon Husted, a Republican, at 32 percent. That lines up with the other lenses. Markets give Brown a 55 percent chance to win against 46 percent for Husted, and recent polling puts Brown at 51 percent of the vote to Husted's 44 percent. Libertarian Bill Redpath still shows up in the chatbot answers at 16 percent, a share the markets and polls do not register at all, while independent Greg Levy sits at just 4 percent. Nothing moved much today. The race has held steady for over a week, with all three lenses still pointing the same direction, 45 days before the November 3 general election.

    Written by Claude Sonnet 5, checked by GPT-5.5

  4. Alaska State Senate District O Race 2026

    Alaska chatbots still put Bauer ahead of Sheldon, more than a month after voters ranked Sheldon higher in District O.

    All three State Senate District O candidates advanced to November after Alaska's top four primary on August 18, since all three finished within the top four. Voters put Republican incumbent George Rauscher first at 38.8 percent, then Sheldon and Bauer essentially tied at 30.6 percent each. Today, chatbots still pick Rauscher most often, at 47.3 percent, matching the actual winner. But they rank Democrat Peter Bauer second, at 29.9 percent, ahead of Republican Ryan Sheldon at 15 percent, even though voters placed Sheldon just ahead of Bauer on primary night. That inversion has held for weeks. There is no live prediction market or recent poll to check the chatbots against for the general election.

    Written by Claude Sonnet 5, checked by GPT-5.5

  5. Texas U.S. Senate Race 2026

    Texas chatbots, markets and polls all keep pointing to Talarico in the Senate race

    Chatbots asked by Texas voters about the U.S. Senate race still pick Democrat James Talarico most often, at 49.1 percent, ahead of Republican Ken Paxton at 37.5 percent. Betting markets agree, giving Talarico a 59.5 percent chance to win against Paxton's 41.5 percent. Recent polling from Texas Public Opinion Research, the University of Texas and Emerson College, last updated August 24, also puts Talarico ahead, 45.3 percent to 42.7 percent. Libertarian Ted Brown draws 10.9 percent from the chatbots but is not tracked by markets or polls. Nothing moved much today, so the race remains where it has been for over a week, with all three measures reading the same way.

    Written by Claude Sonnet 5, checked by GPT-5.5

  6. Iowa U.S. Senate Race 2026

    Iowa chatbots still split almost evenly on the Senate race while betting markets keep favoring Hinson.

    Voters in Iowa asking chatbots about the U.S. Senate race get a near coin flip. Republican Ashley Hinson gets picked about 42.2 percent of the time, Democrat Josh Turek about 43.4 percent, with Libertarian Thomas Laehn far behind at 13.3 percent. Betting markets see it differently, giving Hinson a 61.5 percent chance to win against Turek's 39 percent, a gap of roughly 19 points that has held steady for over a week. Recent polling, last updated September 1, has the two majors essentially tied, Hinson at 48.7 percent and Turek at 48.3 percent. Nothing moved enough today to call it new, the chatbots' even split and the markets' lean toward Hinson are both continuing trends 45 days before the November 3 general election.

    Written by Claude Sonnet 5, checked by GPT-5.5

  7. Maine U.S. Senate Race 2026

    Maine voters asking about Senate still get Collins from the chatbots, even as bettors keep favoring Jackson

    Ask an AI chatbot who should win Maine's Senate race and it still points to incumbent Susan Collins, picked about 54 percent of the time. The betting markets disagree, giving Democrat Troy Jackson roughly a 70 percent chance to win, compared with about 31 percent for Collins. Recent polling is closer, putting Jackson at 47.3 percent and Collins at 45.7 percent, though that poll is over three weeks old. This gap between what the chatbots say and what the markets expect has held steady for more than a week, with neither side moving much today. Nothing new shifted the picture, the chatbots and the markets are simply still reading this race in opposite directions, 45 days before the general election.

    Written by Claude Sonnet 5, checked by GPT-5.5

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

    “Alaska chatbots still favor Sullivan for Senate, and the betting markets just widened their lean toward Peltola” (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 forbidden parentheses in Sullivan (R).

    Written by Claude Sonnet 5, disputed by GPT-5.5

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

    “Alaska's chatbots keep backing Begich for the House seat, and nothing new moved today” (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 forbidden parentheses in Begich (R).

    Written by Claude Sonnet 5, disputed by GPT-5.5

Theme