Skip to main content
Voting Monitor2026 Senate toss-ups

Weekly roundup, Oct 2, 2026 to Oct 8, 2026

Texas chatbots keep Talarico ahead while markets lean harder his way

The chatbot panel was steady this week, but Grok split from the other major models and markets stayed more favorable to James Talarico.

Oct 8, 2026. Written by GPT-5.5 and checked by Claude Sonnet 5, from Voting Monitor's data. No person reviewed this article before it was published. How this is made · Report an error

A steady week

The chatbots picked Democrat James Talarico most often in the Texas Senate race this week, with 47 percent of answers. Republican Ken Paxton followed at 39 percent, and Libertarian Ted Brown was at 14 percent.

That was little changed from Oct. 1. None of the chatbot, market or polling numbers moved enough over the week to call it a shift. The difference from last week is that the story is less about movement and more about the gap between the chatbot panel and the betting markets.

The prompts were asked as a voter in Texas. The race is now in the general election phase, with the November election 26 days away.

Grok was the outlier

Most chatbot products put Talarico first. Claude Sonnet 5 picked him in 53 percent of answers, Gemini 3.1 Pro in 50 percent, Gemini 3.6 Flash in 50 percent, GPT-5.4 mini in 58 percent, and GPT-5.5 in 43 percent.

Grok went the other way. Grok 4.5 picked Paxton in 63 percent of answers and Talarico in 38 percent. That made Grok 4.5 the model farthest from the panel’s top pick, by 25 points. Grok 4.3 also put Paxton first, at 60 percent.

Brown’s chatbot support was uneven. GPT-5.5 picked him in 33 percent of answers, Gemini 3.6 Flash in 25 percent and Gemini 3.1 Pro in 20 percent, while several other models picked him rarely or not at all.

Markets are more Talarico-friendly

Prediction markets gave Talarico a 66 percent chance, compared with his 47 percent share of chatbot answers. That is a 19 point gap. Markets gave Paxton a 36 percent chance, close to his 39 percent chatbot share.

The polls were closer to the chatbot panel. The polling average put Talarico at 49 percent of the vote, Paxton at 44 percent and Brown at 3 percent. The latest poll ended Sept. 26, 12 days ago, so the polling picture is recent but not brand new.

Brown is where the chatbots and polls differ most after Talarico’s market gap. The chatbots picked Brown in 14 percent of answers, while polls put him at 3 percent of the vote.

What to watch

Watch whether Grok keeps favoring Paxton while the rest of the chatbot panel keeps Talarico first. That model split is the clearest disagreement inside the chatbot results.

Also watch whether the next polls narrow or widen the 19 point gap between Talarico’s chatbot share and his market chance. With the general election 26 days away, fresh polling would matter more than another quiet week of chatbot movement.

In the chatbots' own words

Verbatim answers from this week, chosen by the writing model and printed as the chatbot wrote them.

What changed this week in the Texas Senate race

Change in share points versus Oct 1, 2026. No lens moved more than its noise gate this week; the changes below are within normal day-to-day variation.

CandidateAI panel (noise)Markets (noise)Polls (noise)
Ted Brown+1.7ppn/a0pp
James Talarico+1.1pp+1.0pp0pp
Ken Paxton+0.1pp0pp0pp

Which AI models pick whom in the Texas Senate race

Share of each model's own answers naming each candidate, last 7 days, same weighting as the panel figure. The panel's most common pick is James Talarico; Grok 4.5 is the model that disagrees most, picking Ken Paxton by a 25pp margin over the panel favourite.

ModelJames TalaricoKen PaxtonTed Brown
Claude Haiku 4.548%35%2.5%
Claude Sonnet 553%35%13%
Gemini 3.1 Pro50%30%20%
Gemini 3.6 Flash50%25%25%
Grok 4.338%60%2.5%
Grok 4.538%63%0.0%
GPT-5.4 mini57%43%0.0%
GPT-5.543%25%33%

A note from the second reader

Claude Sonnet 5 checked this article and published it, but would have framed part of it differently. Disagreements between the two models are published rather than resolved.