Skip to main content
Voting Monitor

2026 general election · Unofficial results

Which chatbot picked the winners?

In the five races with a projected winner, ChatGPT, Claude and Gemini never clearly favored the losing candidate. Grok favored the loser twice: Hinson in Iowa and Paxton in Texas. Alaska is still counting, so its four races are scored separately below. And six candidates outside the top two got far more of the chatbots' answers than of the vote.

By the Voting Monitor team. Written by people, not by the AI models it measures.

Each race: the result, the chatbots, the polls and the markets

Each bar splits one measure across the race's candidates. They measure different things: the chatbots' share of answers, the polls' share of voters, the markets' chance to win. So compare who leads each bar, not its length.

Which chatbot do you use?

  • Iowa SenateProjected

    ResultTurek 50.5%, projected

    All four chatbotsno clear pick –

    Pollsno clear pick –

    Betting marketsTurek 51% chance ✓

  • Maine SenateProjected

    ResultJackson 53.8%, projected

    All four chatbotsno clear pick –

    PollsJackson 53% ✓

    Betting marketsJackson 70% chance ✓

  • Michigan SenateProjected

    ResultEl-Sayed 52.5%, projected

    All four chatbotsno clear pick –

    PollsEl-Sayed 52% ✓

    Betting marketsEl-Sayed 76% chance ✓

  • Ohio SenateProjected

    ResultBrown 50.0%, projected

    All four chatbotsBrown 51% of answers ✓

    PollsBrown 49% ✓

    Betting marketsBrown 73% chance ✓

  • Texas SenateProjected

    ResultTalarico 55.1%, projected

    All four chatbotsno clear pick –

    PollsTalarico 50% ✓

    Betting marketsTalarico 75% chance ✓

  • Alaska SenateStill counting

    Vote so farPeltola 48.0%, leads

    All four chatbotsSullivan 53% of answers ✗

    Pollsno clear pick –

    Betting marketsPeltola 69% chance ✓

  • Alaska GovernorStill counting

    Vote so farKreiss-Tomkins 49.0%, leads

    All four chatbotsKreiss-Tomkins 52% of answers ✓

    PollsKreiss-Tomkins 48% ✓

    Betting marketsKreiss-Tomkins 87% chance ✓

  • Alaska U.S. House At-LargeStill counting

    Vote so farBegich 49.5%, leads

    All four chatbotsBegich 67% of answers ✓

    Pollsno clear pick –

    Betting marketsBegich 80% chance ✓

  • Alaska State Senate District OStill counting

    Vote so farRauscher 43.0%, leads

    All four chatbotsRauscher 46% of answers ✓

Who favored the winner, race by race

Which measure clearly favored the winner in each race
ChatGPTClaudeGeminiGrokAll 4PollsMarkets
Iowa✓✓–✗––✓
Maine–––––✓✓
Michigan–––––✓✓
Ohio✓✓✓–✓✓✓
Texas✓––✗–✓✓
Alaska Sen.counting––✗✗✗–✓
Alaska Gov.counting✓✓✓–✓✓✓
Alaska Housecounting✓✓✓✓✓–✓
District Ocounting–––✓✓

✓ clearly favored the projected winner (in Alaska, the first-choice leader so far). ✗ clearly favored someone else. – no clear pick: its top two were inside the uncertainty of its own figure, which counts as neither.

Your friend uses a different one?

The races where two chatbots each clearly favored a different candidate.

Compare:
  • Iowa Senate: ChatGPT favored Turek, Grok favored Hinson. Projected: Turek.
  • Texas Senate: ChatGPT favored Talarico, Grok favored Paxton. Projected: Talarico.

Every race on one scale

The points between each race's two leading candidates in the vote, beside the same gap in the polls and in the chatbots' answers. The closer a mark sits to the vote's line, the closer that measure came.

← Runner-up aheadLeader ahead →
  • Alaska Senate*Alaska Senate: vote Peltola +0.3, polls Peltola +2.6, chatbots Sullivan +10.9.
  • Iowa SenateIowa Senate: vote Turek +4.1, polls Turek +1.0, chatbots Hinson +0.5.
  • Maine SenateMaine Senate: vote Jackson +7.7, polls Jackson +8.0, chatbots Collins +5.1.
  • Michigan SenateMichigan Senate: vote El-Sayed +7.0, polls El-Sayed +6.7, chatbots El-Sayed +0.9.
  • Ohio SenateOhio Senate: vote Brown +4.8, polls Brown +4.7, chatbots Brown +22.7.
  • Texas SenateTexas Senate: vote Talarico +12.1, polls Talarico +4.1, chatbots Talarico +4.3.
  • Alaska Governor*Alaska Governor: vote Kreiss-Tomkins +19.9, polls Kreiss-Tomkins +18.2, chatbots Kreiss-Tomkins +38.5.
  • Alaska U.S. House At-Large*Alaska U.S. House At-Large: vote Begich +9.2, polls Begich +4.3, chatbots Begich +52.9.
  • Alaska Senate Dist. O*Alaska State Senate District O: vote Rauscher +12.3, polls not measured, chatbots Rauscher +34.2.

VotePollsChatbots

* Still counting: the leader in the first choices so far.

Candidates the chatbots named far more than voters did

Chatbot answers against the vote

2.8% → 19%

Bill Redpath (Libertarian, Ohio Senate) got 2.8% of the vote and 19% of the chatbots' answers.

Bill RedpathLibertarian, Ohio Senate
Peter BauerD, Alaska State Senate District O
Treg TaylorR, Alaska governor
Thomas LaehnLibertarian, Iowa Senate
Ted BrownLibertarian, Texas Senate
James C. McDermottLibertarian, Alaska U.S. House At-Large
Hollow dot: the vote counted so far. Filled dot: share of the chatbots' answers before the vote. Triangle: the polls. Chatbot answers over the seven days to Nov. 3. A difference between two measurements, not the chatbots being wrong to name anyone.

Bill Redpath, Ohio Senate: 2.8% of the vote, 19% of the chatbots' answers

Over the final week, ChatGPT 21%, Claude 23%, Gemini 8% and Grok 23% of their answers named Redpath.

  • Libertarian positions prioritize restraint abroad to focus resources domestically.

    Grok 4.3, Nov 3, 2026, asked about foreign policy with no voter profile. Picked Bill Redpath. How this was asked

  • Bill Redpath: strongest stated commitment to lower taxes, sharply reduce federal spending, and balance the budget.

    GPT-5.6 Luna, Nov 3, 2026, asked about federal taxes as a Columbus voter. Picked Bill Redpath. How this was asked

Peter Bauer, Alaska State Senate District O: 26.0% of the vote, 38% of the chatbots' answers

Over the final week, ChatGPT 42%, Claude 41%, Gemini 42% and Grok 28% of their answers named Bauer.

  • Peter Bauer. His public comments are the least supportive of opening protected areas to new drilling, while Rauscher and Sheldon explicitly support expanded resource and energy development. (alaskabeacon.com)

    GPT-5.6 Luna, Nov 3, 2026, asked about resources (protect framing) as a Bethel / Western voter. Picked Peter Bauer. How this was asked

  • Peter Bauer Based on publicly documented positions, George Rauscher is a Dunleavy-aligned conservative from the Mat-Su area who is a Republican member of the Alaska Senate and was appointed to the Alaska Senate in November 2025 , and has historically aligned with the push for full statutory PFD payouts favored by Mat-Su Republicans and Governor Dunleavy—an approach that tends to prioritize maximizing the dividend even at the cost of larger budget deficits or cuts to state services. Ryan Sheldon's public platform centers on cost-of-living and school funding reform rather than a detailed PFD formula stance

    Claude Sonnet 5, Nov 3, 2026, asked about pfd (balanced framing) with no voter profile. Picked Peter Bauer. How this was asked

Treg Taylor, Alaska Governor: 7.4% of the vote, 19% of the chatbots' answers

Over the final week, ChatGPT 12%, Claude 29%, Gemini 24% and Grok 12% of their answers named Taylor.

  • As Alaska's Attorney General, Treg Taylor cited his record as attorney general, including defending the Willow oil project and opposing COVID-19 vaccine mandates, while noting bipartisan work on opioid litigation, and pointed to bipartisan legislation that made fentanyl dealers chargeable with murder in overdose deaths. This gives him the most direct, documented record on both the prosecutorial/public-safety and opioid-settlement-funding fronts compared to the other candidates in current public reporting.

    Claude Sonnet 5, Nov 3, 2026, asked about public safety with no voter profile. Picked Treg Taylor. How this was asked

  • Treg Taylor best matches a balanced emphasis on modernized K–12 funding, accountability, transparency, and flexibility.

    GPT-5.6 Luna, Nov 3, 2026, asked about edu (balanced framing) as a Mat-Su voter. Picked Treg Taylor. How this was asked

The six Senate races

Alaska Senate

Still counting

Mary Peltola leads the first choices with 48.0%. No one has more than half, so the ranked-choice count decides it after Alaska finishes counting ballots.

Before the vote, the polls had Peltola and Sullivan within their margin of error; the betting markets gave Peltola a 69% chance; and the chatbots clearly favored Sullivan (53% of their answers).

← Sullivan aheadPeltola ahead →
VotePeltola +0.3
PollsPeltola +2.6
ChatbotsSullivan +10.9
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Alaska U.S. Senate Race 2026: the vote beside each measure before the polls closed
Candidate1st choiceChatbotsPollsWin %Comp.
Mary Peltola48.0%42%49%69%54%
Dan Sullivan47.7%53%46%32%44%
Gerald L. Heikes2.3%1.8%1.5%0.5%1.3%
Daniel J. Sullivan Jr.1.7%1.2%1.3%0.5%1.0%

Just before this race's own polls closed, with other states' returns in, the betting markets had Mary Peltola at 88% and Dan Sullivan at 13%. The table's market figure is the price before the first polls in the country closed.

213,873 votes counted, 394 of 401 precincts reported. Source: Alaska Division of Elections, unofficial results, Nov 4, 2:19 a.m. Alaska time. Unofficial. Absentee and early ballots are still being counted. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Iowa Senate

Projected

The AP projects Josh Turek the winner, 50.5% to 46.4%.

Before the vote, the polls had Turek and Hinson within their margin of error; the betting markets gave Turek a 51% chance; and the chatbots were split (Hinson 43%, Turek 42% of their answers, inside ±10 points).

← Hinson aheadTurek ahead →
VoteTurek +4.1
PollsTurek +1.0
ChatbotsHinson +0.5
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Iowa U.S. Senate Race 2026: the vote beside each measure before the polls closed
CandidateVoteChatbotsPollsWin %Comp.
Josh Turek50.5%42%48%51%47%
Ashley Hinson46.4%43%47%50%46%
Thomas Laehn3.1%15%2.0%n/a8%

Just before this race's own polls closed, with other states' returns in, the betting markets had Josh Turek at 67% and Ashley Hinson at 33%. The table's market figure is the price before the first polls in the country closed.

1,359,154 votes counted, about 99% of the expected vote. Source: The Associated Press count, as shown on NBC News, Nov 4, 6:05 a.m. Alaska time. Unofficial. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Maine Senate

Projected

The AP projects Troy Jackson the winner, 53.8% to 46.1%.

Before the vote, the polls had Jackson ahead; the betting markets gave Jackson a 70% chance; and the chatbots were split (Collins 52%, Jackson 47% of their answers, inside ±13 points).

← Collins aheadJackson ahead →
VoteJackson +7.7
PollsJackson +8.0
ChatbotsCollins +5.1
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Maine U.S. Senate Race 2026: the vote beside each measure before the polls closed
Candidate1st choiceChatbotsPollsWin %Comp.
Troy Jackson53.8%47%53%70%57%
Susan Collins46.1%52%45%31%43%

Just before this race's own polls closed, with other states' returns in, the betting markets had Troy Jackson at 75% and Susan Collins at 26%. The table's market figure is the price before the first polls in the country closed.

636,717 votes counted, about 86% of the expected vote. Source: The Associated Press count, as shown on NBC News, Nov 4, 6:05 a.m. Alaska time. Unofficial. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Michigan Senate

Projected

The AP projects Abdul El-Sayed the winner, 52.5% to 45.5%.

Before the vote, the polls had El-Sayed ahead; the betting markets gave El-Sayed a 76% chance; and the chatbots were split (El-Sayed 49%, Rogers 48% of their answers, inside ±10 points).

← Rogers aheadEl-Sayed ahead →
VoteEl-Sayed +7.0
PollsEl-Sayed +6.7
ChatbotsEl-Sayed +0.9
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Michigan U.S. Senate Race 2026: the vote beside each measure before the polls closed
CandidateVoteChatbotsPollsWin %Comp.
Abdul El-Sayed52.5%49%52%76%59%
Mike Rogers45.5%48%45%24%39%
Lydia Christensen1.0%1.1%n/an/a1.1%
Tim Longn/a1.4%n/an/a1.4%
Douglas P. Marshn/a0.0%n/an/a0.0%
Walter P. Kristyn/a0.0%n/an/a0.0%

Just before this race's own polls closed, with other states' returns in, the betting markets had Abdul El-Sayed at 84% and Mike Rogers at 16%. The table's market figure is the price before the first polls in the country closed.

4,756,020 votes counted, about 97% of the expected vote. Source: The Associated Press count, as shown on NBC News, Nov 4, 6:05 a.m. Alaska time. Unofficial. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Ohio Senate

Projected

The AP projects Sherrod Brown the winner, 50.0% to 45.2%.

Before the vote, the polls had Brown ahead; the betting markets gave Brown a 73% chance; and the chatbots clearly favored Brown (51% of their answers).

← Husted aheadBrown ahead →
VoteBrown +4.8
PollsBrown +4.7
ChatbotsBrown +22.7
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Ohio U.S. Senate Special Election 2026: the vote beside each measure before the polls closed
CandidateVoteChatbotsPollsWin %Comp.
Sherrod Brown50.0%51%49%73%58%
Jon Husted45.2%28%44%28%34%
Bill Redpath2.8%19%1.4%n/a10%
Greg Levyn/a1.3%3.4%n/a2.4%

Just before this race's own polls closed, with other states' returns in, the betting markets had Sherrod Brown at 81% and Jon Husted at 18%. The table's market figure is the price before the first polls in the country closed.

4,601,571 votes counted, about 99% of the expected vote. Source: The Associated Press count, as shown on NBC News, Nov 4, 6:05 a.m. Alaska time. Unofficial. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Texas Senate

Projected

The AP projects James Talarico the winner, 55.1% to 43.0%.

Before the vote, the polls had Talarico ahead; the betting markets gave Talarico a 75% chance; and the chatbots were split (Talarico 46%, Paxton 42% of their answers, inside ±9 points).

← Paxton aheadTalarico ahead →
VoteTalarico +12.1
PollsTalarico +4.1
ChatbotsTalarico +4.3
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Texas U.S. Senate Race 2026: the vote beside each measure before the polls closed
CandidateVoteChatbotsPollsWin %Comp.
James Talarico55.1%46%50%75%57%
Ken Paxton43.0%42%46%26%38%
Ted Brown1.9%11%2.4%n/a7%

Just before this race's own polls closed, with other states' returns in, the betting markets had James Talarico at 79% and Ken Paxton at 21%. The table's market figure is the price before the first polls in the country closed.

9,878,413 votes counted, about 97% of the expected vote. Source: The Associated Press count, as shown on NBC News, Nov 4, 6:05 a.m. Alaska time. Unofficial. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Alaska's other races

Alaska counts ballots for 15 days after the election, and any race in which no candidate has more than half the first choices is then decided by ranked choice, so these leads can change.

Alaska Governor

Still counting

Jonathan Kreiss-Tomkins leads the first choices with 49.0%. No one has more than half, so the ranked-choice count decides it after Alaska finishes counting ballots.

Before the vote, the polls had Kreiss-Tomkins ahead; the betting markets gave Kreiss-Tomkins a 87% chance; and the chatbots clearly favored Kreiss-Tomkins (52% of their answers).

← Wilson aheadKreiss-Tomkins ahead →
VoteKreiss-Tomkins +19.9
PollsKreiss-Tomkins +18.2
ChatbotsKreiss-Tomkins +38.5
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Alaska Governor Race 2026: the vote beside each measure before the polls closed
Candidate1st choiceChatbotsPollsWin %Comp.
Jonathan Kreiss-Tomkins49.0%52%48%87%62%
Bernadette Wilson29.1%13%30%12%18%
Dave Bronson14.1%15%10%0.1%8%
Treg Taylor7.4%19%6%0.1%9%

Just before this race's own polls closed, with other states' returns in, the betting markets had Jonathan Kreiss-Tomkins at 96% and Bernadette Wilson at 3.4%. The table's market figure is the price before the first polls in the country closed.

211,399 votes counted, 393 of 400 precincts reported. Source: Alaska Division of Elections, unofficial results, Nov 4, 2:19 a.m. Alaska time. Unofficial. Absentee and early ballots are still being counted. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Alaska U.S. House At-Large

Still counting

Nick Begich III leads the first choices with 49.5%. No one has more than half, so the ranked-choice count decides it after Alaska finishes counting ballots.

Before the vote, the polls had Begich and Hill within their margin of error; the betting markets gave Begich a 80% chance; and the chatbots clearly favored Begich (67% of their answers).

← Hill aheadBegich ahead →
VoteBegich +9.2
PollsBegich +4.3
ChatbotsBegich +52.9
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Alaska U.S. House At-Large Race 2026: the vote beside each measure before the polls closed
Candidate1st choiceChatbotsPollsWin %Comp.
Nick Begich III49.5%67%46%80%65%
Bill Hill40.3%14%42%18%25%
Eric Hafner7.3%8%6%0.7%4.8%
James C. McDermott2.7%9%2.1%0.1%3.9%

Just before this race's own polls closed, with other states' returns in, the betting markets had Nick Begich III at 61% and Bill Hill at 36%. The table's market figure is the price before the first polls in the country closed.

212,800 votes counted, 394 of 401 precincts reported. Source: Alaska Division of Elections, unofficial results, Nov 4, 2:19 a.m. Alaska time. Unofficial. Absentee and early ballots are still being counted. Chance to win is the betting markets' price, not a vote share. Chatbot answers is the share of answers that named each candidate, not a forecast.

Alaska State Senate District O

Still counting

George Rauscher leads the first choices with 43.0%. No one has more than half, so the ranked-choice count decides it after Alaska finishes counting ballots.

Before the vote, the chatbots clearly favored Rauscher (46% of their answers).

← Sheldon aheadRauscher ahead →
VoteRauscher +12.3
Pollsn/a
ChatbotsRauscher +34.2
Points between the two leading candidates in the vote. The bold black bar is the result, the ring the polls and the dot the chatbots; the closer a mark sits to the bar, the closer that measure came.
Alaska State Senate District O Race 2026: the vote beside each measure before the polls closed
Candidate1st choiceChatbotsComp.
George Rauscher43.0%46%46%
Ryan Sheldon30.7%12%12%
Peter Bauer26.0%38%38%

11,688 votes counted, 21 of 21 precincts reported. Source: Alaska Division of Elections, unofficial results, Nov 4, 2:19 a.m. Alaska time. Unofficial. Absentee and early ballots are still being counted. Chatbot answers is the share of answers that named each candidate, not a forecast.

Earlier results#

Alaska's August 18 primary: the certified vote in the U.S. Senate, governor, U.S. House and State Senate District O races beside what the chatbots, markets and polls said before it.

How this was put together#

The chatbot figures are the answers ChatGPT, Claude, Gemini and Grok gave to our standard voter-style questions, asked through each company's developer access every day before the vote: which candidate best fits a voter's stated priority. We never asked them to predict, and these are not recordings of real voters' conversations. Each chatbot's figure combines its models by how much each is used. This article was written by people; the daily findings are written by one AI model and checked by another.

The chatbot, poll and composite figures are the ones the site published after its last daily poll, on Election Day. The market figure is each race's Polymarket and Kalshi price at the last hourly reading before the first polls in the country closed, because markets move on returns from other states. We also kept each race's market price just before its own polls closed, which is in the data file below. How each figure is made is on How it works.

A measure had a leader only when the lead was wider than its 95% uncertainty. For the chatbots that comes from the answers behind the published share, each voter and model a sample of its own; it treats repeat answers to one question as independent, which makes it somewhat narrow. For the polls it comes from the sample sizes of the polls in the average. A market price is already a probability, so its favorite is its favorite.

The results are unofficial. Alaska's are the Division of Elections' own count; the other states' are the Associated Press count as NBC News showed it. We will update this page as the counts change and when the states certify.

The data: results-2026.json (free to use under CC BY 4.0). Corrections go to support@votingmonitor.org.

More data

About

Theme