2026 election · Unofficial results

ChatGPT and Claude leaned toward every winner

In the five races called so far, ChatGPT and Claude leaned toward the winner every time, clearly in three and two. Grok leaned toward the loser in four, clearly in Iowa and Texas. The polls came closer on the margins.

See Grok's two misses →See every race →

Which measure clearly favored the winner in each race
	Iowa	Maine	Michigan	Ohio	Texas	Alaska Sen.still counting	Called	Alaska
ChatGPT	✓	✓	✓	✓	✓	✗	3 ✓ 0 ✗leaned 2 ✓ 0 ✗	0 ✓ 0 ✗leaned 0 ✓ 1 ✗
Claude	✓	✓	✓	✓	✓	✗	2 ✓ 0 ✗leaned 3 ✓ 0 ✗	0 ✓ 0 ✗leaned 0 ✓ 1 ✗
Gemini	✗	✗	✓	✓	✓	✗	1 ✓ 0 ✗leaned 2 ✓ 2 ✗	0 ✓ 1 ✗
Grok	✗	✗	✗	✓	✗	✗	0 ✓ 2 ✗leaned 1 ✓ 2 ✗	0 ✓ 1 ✗
All four combined	✗	✗	✓	✓	✓	✗	1 ✓ 0 ✗leaned 2 ✓ 2 ✗	0 ✓ 1 ✗
Polls	✓	✓	✓	✓	✓	✓	4 ✓ 0 ✗leaned 1 ✓ 0 ✗	0 ✓ 0 ✗leaned 1 ✓ 0 ✗
Betting markets	✓	✓	✓	✓	✓	✓	4 ✓ 0 ✗leaned 1 ✓ 0 ✗	1 ✓ 0 ✗

Which measure clearly favored the winner in each race
	ChatGPT	Claude	Gemini	Grok	All 4	Polls	Markets
Iowa	✓	✓	✗	✗	✗	✓	✓
Maine	✓	✓	✗	✗	✗	✓	✓
Michigan	✓	✓	✓	✗	✓	✓	✓
Ohio	✓	✓	✓	✓	✓	✓	✓
Texas	✓	✓	✓	✗	✓	✓	✓
Still counting, not in the totals
Alaska Sen.	✗	✗	✗	✗	✗	✓	✓

✓ ✗: a clear call. Faint ✓ or ✗: leaned that way, within its margin of error, so not counted. How we scored it

What the answers show#

Every day we put the same voter questions to ChatGPT, Claude, Gemini and Grok. Here is what the last 30 days of answers say about what a voter is told.

Same question, four chatbots

42%

of the time, ChatGPT, Claude, Gemini and Grok don't all name the same candidate.

How often the four disagree, by race

All races, 42%

Share of identical questions (same day, voter and issue) where at least one chatbot named someone else, last 30 days, 2,354 questions. Races with more candidates leave more room to disagree.

Texas Senate · one question, four chatbots · Nov 2, 2026

All four named Ken Paxton.

“The issue I care about most is U.S. immigration policy and the border. Considering the priority of tightening border enforcement and reducing unlawful immigration, which candidate's record and policy positions most closely align with that priority?”

ChatGPT
Ken Paxton
“C — Ken Paxton; his record and stated policies most strongly emphasize border enforcement and reducing unlawful immigration. (fox4news.com)”
Claude
Ken Paxton
“Based on the public record, Ken Paxton's positions most closely align with a priority of tightening border enforcement and reducing unlawful immigration. C) Ken Paxton — As Texas Attorney General, Paxton advocates aggressive enforcement-first border and immigration policies, including criminalizing…”
Gemini
Ken Paxton
“B) Ken Paxton”
Grok
Ken Paxton
Answered with the letter of its choice and nothing more.

Answers are word for word from the most-used model behind each chatbot's free version, with the candidates listed in a random order. The question is chosen by a fixed rule (the race's most-asked issue, the latest day all four answered), not for the answers. A pick is not an endorsement.

Show the issue chart: what decides the answer

What decides the answer

The issue.

A voter who puts school funding or immigration, more permissive first is pointed to the Democrat by every chatbot; one who puts veterans and the military or immigration, stricter enforcement first, to the Republican. On federal taxes, the picks go to the Republican or, in Iowa, Ohio and Texas, often to the Libertarian. The chatbots part ways on the issues in between.

ChatGPTClaudeGeminiGrok

← Points to the DemocratPoints to the Republican →

School funding

Immigration, more permissive

Health careoften an independent or minor-party candidate in Alaska

Native subsistenceoften an independent or minor-party candidate in Alaska

Fisheriesoften an independent or minor-party candidate in Alaska

Permanent Fund dividend

Cost of living

Education

Immigration, balanced

Resource development

Balanced foreign policyoften Libertarian in Iowa, Ohio and Texas

Housing

Foreign policy

Federal infrastructure

Public safety

Immigration, stricter enforcement

Federal taxesoften Libertarian in Iowa, Ohio and Texas

Veterans and the military

Every race combined, last 30 days. Each dot is one chatbot; the bar joins the four. The ends of the scale mean every pick went to one party's candidate. The scale counts Republican and Democratic picks only; where an independent or minor-party candidate is the pick of at least two chatbots in a race, the row says so.

More in the full analysis: six Senate races compared, how each chatbot behaves, and each race's page below.

Each chatbot counts once, its models split by how much each is used. Margins compare the Republican and Democratic candidates; a pick is a match to one voter's stated priority, not an endorsement or a forecast, and the polls are not the truth: a gap says the two disagree. Through Nov 3, 2026.

The final daily finding

Final FindingNov 4, 2026 edition, Alaska time

From answers collected Tuesday, November 3.

Texas Senate markets moved toward Talarico on Election Day while chatbots stayed closely split

Texas voters asking chatbots about the U.S. Senate race got a close call, with chatbots picking Democrat James Talarico 46.4 percent of the time and Ken Paxton (R) 42.1 percent. Betting markets moved 3.0 points toward Talarico and now give him a 74.5 percent chance to win. Polls put Talarico at 49.6 percent of the vote and Paxton at 45.5 percent. Grok 4.5, one of Grok's models, favors Paxton by 30.0 points.

Summary byGPT-5.5·checked byClaude Sonnet 5

Explore the Texas Senate race →

Every daily finding →

Why this matters

Same ballot. Same question. Different AI answers. About half of U.S. adults use AI chatbots, and 8% say they are using them for election information this year; 4 in 10 of those want help deciding how to vote. In a controlled study, a short conversation with a chatbot moved voters’ candidate preferences more than video ads typically do. This site tracks what those voters are told.

Sources: Pew Research Center, Oct. 2026; Nature, Dec. 2025.