See the full Winning the AI Vote report here.
People will ask large language models (LLMs) just about anything: how to fix a leaky faucet, where to eat on vacation, even who to vote for.
About 650,000 prompts a week were about Republicans or Democrats in August, according to data from EverPanel’s more than 150 million real user conversations. About 40,000 prompts a week focused on November’s midterm elections in particular.
As of: July 15, 2026
The vast majority (70-85%) of those prompting about Republicans, Democrats or the midterm elections did so on ChatGPT. Gemini consistently came in 2nd with a meaningful share, while Copilot and Perplexity combined for a tiny fraction of these prompt volumes.
What models won’t say
Models learn in post-training what a helpful response looks like and what they should and should not say to users. This is an essential step to prevent misuse, like asking an LLM for help in committing a crime, or harm, especially when users share mental health struggles with an LLM.
Sometimes, the differences in the level of caution in models’ responses can be quite noticeable model-to-model. For example, we asked models in March where college students should go for spring break. Many responses recommended locations in Mexico, despite upheaval there in late February.
However, Gemini mentioned the events and advised caution and reliance on U.S. State Department travel advisories, while Copilot not only failed to note any of February’s events but recommended Puerto Vallarta (an area particularly affected) when asked for the “safest international destinations.”
Like would-be spring breakers, future voters asking LLMs about candidates and elections are highly likely to take action based on the information LLMs give them. How cautious were LLMs on this topic?
Unlike with our spring break results, we found that all models were very constrained and careful. None of the models we prompted - ChatGPT, Copilot, Gemini, Google AI Mode, Google AI Overview and Perplexity - would:
- Endorse a candidate or give a name when asked “who should I vote for”
- Forecast a winner
- Offer advice on how to win an election
Models often included standard disclaimers like:
- ChatGPT: “I can’t tell you who you should vote for, but I can give you a neutral comparison so you can decide based on your priorities.”
- Gemini: “Deciding who to vote for is a personal choice that depends on your individual values, policy priorities, and views on the candidates.”
- Google AI Mode: “Deciding who to vote for depends on your personal political values, party alignment, and the key policy issues that matter most to you.”
Models rely on news coverage when discussing elections
When the models responded to our prompts about what the top voter issues are, what candidates’ platforms say and how candidates and parties compare in specific issues, models went to earned domains, especially news outlets, far more than we typically see.
For example, about 86% of ChatGPT’s sources on prompts about the midterm elections. For all Evertune prompts in July, by contrast, only 41% of sources were earned domains. This suggests model makers tilted models’ sourcing behaviors toward trusted sources on election topics.
Dive deeper
For many more insights on how LLMs talk about elections and candidates plus a deep dive into what they’re saying about Georgia’s Senate race between Jon Ossoff and Mike Collins, check out our full report with Orchestra here.
Methodology
At Evertune, we track thousands of brands across LLMs by running millions of prompts a day. For this analysis, we ran 63,000 prompts in total across ChatGPT, Copilot, Gemini, Google AI Mode, Google AI Overview and Perplexity in July.
Prompts focused on four areas: top voter issues nationwide, where the Democratic and Republican parties stand on those issues, the top issues in the Georgia Senate race and where Jon Ossoff and Mike Collins stand on those issues.
Because LLMs are probabilistic, each unique prompt was sampled multiple times for statistically significant findings.
Evertune is the first Generative Engine Optimization (GEO) platform built to explore, measure, act and advertise across the entire AI customer journey, connecting brands directly to ChatGPT Ads and programmatic advertising partners like The Trade Desk and Index Exchange. Where most GEO tools sample each prompt once a day, Evertune samples every prompt 100 times per model across 11+ AI models, delivering statistically significant visibility data at half the cost of competing platforms. Evertune's agents do the work for you: a prompt agent mines over 150 million real user conversations to find the exact questions buyers ask, an insights agent tells you what to do next each week, and an ads agent builds a complete ChatGPT campaign around your visibility gaps. From there, Evertune closes the loop with website optimization, data-driven content creation, source-level influence mapping and paid activation through affiliate and programmatic AI retargeting partners. Founded by the team that pioneered programmatic advertising at The Trade Desk, now building the next marketing channel.
