About this snapshot

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

1. Alaska Governor Race 2026

   Chatbots give Thomas Scott Begich far more room than the betting markets do

   Alaska chatbots picked Democrat Thomas Scott Begich 22.2 percent of the time for governor, while markets give him a 1.3 percent chance to win. That is the clearest live split in this race. The same chatbots picked Democrat Jonathan Kreiss Tomkins most often, at 31.7 percent, while markets give him a 60.2 percent chance to win. Polls are older context here, with the latest ending July 1.

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

2. Ohio U.S. Senate Special Election 2026

   Chatbots swung to Husted, while markets and polls still favor Brown

   Chatbots swung toward Republican incumbent Jon Husted, picking him 53.1 percent of the time, up 14.6 points from yesterday. That puts the AI panel opposite the other live yardsticks. Markets give Democrat Sherrod Brown a 56.5 percent chance to win, and polls put Brown at 51.3 percent of the vote, with Husted at 44.0 percent.

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

3. Texas U.S. Senate Race 2026

   Only the chatbots put Ted Brown visibly in the Texas Senate mix

   Chatbots picked Libertarian Ted Brown 18.0 percent of the time for U.S. Senate, while this race has no market or poll number for him in the data. The main lenses still cluster around Democrat James Talarico and Republican Ken Paxton. Chatbots picked Talarico at 40.8 percent and Paxton at 36.8 percent. Markets give each a 50.5 percent chance to win. Polls put Talarico at 45.3 percent of the vote and Paxton at 42.7 percent, based on polling that is 15 days old.

   Written by GPT-5.5, checked by Claude Sonnet 5

4. Maine U.S. Senate Race 2026

   Chatbots still pick Collins while markets favor Jackson for Maine Senate

   Chatbots still picked incumbent Susan Collins (R) far more often than the betting markets did. The AI panel picked Collins 57.7 percent of the time, while markets give her a 29.5 percent chance to win. Democrat Troy Jackson is the mirror image, chatbots picked him 40.3 percent, while markets give him a 70.5 percent chance. Polls are much closer, putting Jackson at 47.3 percent of the vote and Collins at 45.7 percent.

   Written by GPT-5.5, checked by Claude Sonnet 5

5. Iowa U.S. Senate Race 2026

   Iowa Senate chatbots still lean Turek while markets favor Hinson

   Chatbots picked Democrat Josh Turek slightly more often than Republican Ashley Hinson, 49.2 percent to 45.3 percent. The live betting markets read the race the other way, giving Hinson a 65.5 percent chance to win and Turek 34.5 percent. Polls are tighter, putting Hinson at 48.7 percent of the vote and Turek at 48.3 percent. That split is real, but it is not new. The same AI versus markets divide was already visible this week, and none of the three lenses made a clear move since Monday.

   Written by GPT-5.5, checked by Claude Sonnet 5

6. Alaska State Senate District O Race 2026

   Chatbots still pick Rauscher in District O, with no live market or poll to check them

   For the general election field, chatbots picked incumbent George Rauscher (R) 54.1 percent of the time, ahead of Democrat Peter Bauer at 27.3 percent and Republican Ryan Sheldon at 9.8 percent. There is no live prediction market or poll in the data for this race. In the August 18 top four primary, voters put Rauscher first at 38.8 percent, then Sheldon and Bauer at 30.6 percent each. The panel got the top name right but still ranks Bauer over Sheldon. All three advanced, so that primary inversion is history, not a November call.

   Written by GPT-5.5, checked by Claude Sonnet 5

7. Alaska U.S. Senate Race 2026

   Alaska's primary is over. The Senate and House fields are set for November.

   Alaska's U.S. Senate and U.S. House primaries have been certified. Mary Peltola advanced with 49.5 percent of the vote and Dan Sullivan with 41.4 percent in the Senate race. In the House race, Nick Begich III advanced with 44.6 percent, Bill Hill with 32.5 percent, and Matthew Schultz with 8.1 percent. The chatbot panel and betting markets are now describing the general election matchup, not the primary that already happened.

   Written by GPT-5.5, not checked by Claude Sonnet 5

8. Michigan U.S. Senate Race 2026

   Michigan's Senate race split three ways: AI, markets, and polls all disagree on the margin

   Chatbots put Abdul El-Sayed's lead over Mike Rogers at more than 28 points, 61.9 percent to 33.8 percent. The betting markets show a similar gap, 66.0 to 33.5. But recent polls put the race closer, El-Sayed at 48.0 percent and Rogers at 41.7 percent, days_since_latest_poll of 11. The AI panel and markets agree on direction but the polling gap is roughly half the size the other two lenses show.

   Written by GPT-5.5, not checked by Claude Sonnet 5

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

   “Chatbots moved sharply toward Abdul El-Sayed in Michigan Senate answers” (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 phrase 'since Monday' is an imprecise day-count claim. The comparison is actually yesterday-to-today (2026-09-07 to 2026-09-08) per movement_json, and today is a Tuesday, so 'since Monday' overstates the window by one day and is not clearly grounded in election_anchor. Should read 'since yesterday' or 'day over day.'

   Written by GPT-5.5, disputed by Claude Sonnet 5

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

    “Chatbots and older polls both point to Begich in Alaska’s House race” (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: This lead treats AK-AL-HOUSE as an unresolved contest heading toward a future November vote ('The race is now aimed at November, 56 days away') without disclosing that the actual_election_results show this primary is already certified with Begich, Hill, and Schultz advancing. This omits the decided reality check the rules require and reads as a present-tense forecast of a race that has already happened.

    Written by GPT-5.5, disputed by Claude Sonnet 5

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

    “Chatbots still pick Sullivan while markets favor Peltola for Alaska Senate” (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: This lead frames AK-SEN as a live, ongoing horse race using only AI/market/poll numbers, without ever noting that actual_election_results.AK-SEN shows a certified primary result (Peltola 49.5%, Sullivan 41.4%, both advanced). The lead should clarify that these numbers describe expectations for the general election following a completed primary, not an as-yet-undecided contest.

    Written by GPT-5.5, disputed by Claude Sonnet 5