The First GEO Election: 2026 Midterms & AI Search
AI NewsVoters asked AI. Campaigns paid to be the answer.
Key Takeaways:
-
Between 8% and 25% of U.S. adults used AI for 2026 midterm information, depending on the survey.
-
At least 38% of likely voters got election information from AI summaries, and only about half of them realized it.
-
Some campaigns paid for GEO to win AI recommendations.
-
Google largely withheld AI answers on election queries, while ChatGPT and Perplexity answered directly.
-
Chatbots often cited low-quality, AI-built sites – a clear sign that third-party footprint shapes what AI says.
Executive Summary: In the 2026 midterms, between 8% and 25% of U.S. adults asked AI about the election, and some campaigns tried paying their way into AI recommendations. For brands, the most important takeaways are: AI visibility is a premium commodity, and third-party sources shape what AI says about you.
Ahead of the 2026 midterms, a minimum of 8% of U.S. adults used AI chatbots for election information.
8% is roughly 22 million people.
And campaigns priced their attention like ad inventory.
The result was an election in which Generative Engine Optimization (GEO) was openly pitched, priced, and sold to candidates.
More importantly, it was a live stress test of what happens when a wrong answer has a real-world impact – in this case, potentially costing someone their vote.
How did voters use AI in the 2026 midterms?
Based on various surveys leading up to the 2026 midterm election, between 8% and 25% of U.S. adults deliberately used AI engines for election information or candidate research, and up to 40% for at least one election task. Meanwhile, at least 38% of likely voters received election information through AI search summaries – without asking AI directly.
How many Americans asked AI about the midterms?
Based on the above range and the Census data, roughly 22 to 67 million Americans asked AI about the 2026 midterms. The range is wide because different surveys measured different populations at different points in the cycle and/or asked different questions, so their figures are not directly comparable.
| AI Use for 2026 Midterm Election Information, by Survey | |||
| Survey | Finding | Population | Fielded |
| Pew Research Center | 8% used AI chatbots for election info | 5,273 U.S. adults | Oct 12-18, 2026 |
| NCoC / Carter Center (SSRS) | 25% used AI to research candidates; 40% for any of four election tasks | ~1,030 U.S. adults | Oct 15-18, 2026 |
| AAPC Foundation | 1 in 5 plan to use AI to research a race; 38%+ received election info via AI search summaries | 1,568 likely voters | Sep 14-22, 2026 |
| Qlik (Censuswide) | 11% use AI chatbots for election info | 2,000 registered voters | Sep 11-15, 2026 |
| Note: Surveys differ in population (all U.S. adults vs. registered vs. likely voters), timing, and question wording, so their figures aren’t directly comparable. AAPC’s “1 in 5” measures planned use, not actual use. The Qlik survey was sponsored by Qlik; AAPC doesn’t disclose its pollster, survey mode, or margin of error. NCoC’s margin of error is ±3.4 points. | |||
Pew Research Center found that 8% of 5,273 U.S. adults were using AI chatbots to seek out election-related information in mid-October. A Qlik-sponsored Censuswide survey of 2,000 registered voters puts that figure closer to 11%.
Research from the National Conference on Citizenship reported 25% of U.S. adults using AI to learn about Senate, House, or governor candidates, up from 14% just two weeks earlier. Overall, 40% used AI for at least one of four election tasks: researching federal or statewide candidates, researching local candidates and ballot measures, finding out where, when, and how to vote, and verifying political claims, news, or campaign advertisements.
An AAPC Foundation study found that about 20% of voters planned to use AI to research a race on their ballot. On top of that, AAPC also found that many more got AI insights without asking for them: at least 38% of likely voters received election information through AI search summaries – and only about half realized it was AI-generated.
Which voters were most likely to ask AI about the election?
Adults under 50 were nearly three times more likely to seek AI insights about midterms: 11% vs. only 4% of those 50 or older. Granted, this came as no surprise: Pew found that election AI use “largely parallels” the demographics that use chatbots the most.
Our own research points the same way: in a May 2026 ZeroClick Labs survey, we found that AI familiarity predicts trust in government-built AI far more strongly than party affiliation, as well as that the trust deficit that remains isn’t partisan but institutional.
What did voters ask?
Interestingly, among the 8% of Pew-surveyed voters who used chatbots for election information, most asked decision-stage questions:
- 76% looked up where candidates stand on the issues;
- 71% checked the accuracy of claims they had seen elsewhere;
- 40% asked for suggestions on how to make their vote choices.
For brands, this is arguably the most critical piece of insight here: the voters asked AI the same kind of question your buyers ask before they shortlist – “Who’s best for this?”
How did campaigns use GEO in the 2026 midterms?
In 2026, some campaigns began paying for GEO services to make chatbots like ChatGPT, Claude, and Gemini recommend their candidate. Effectively, they bought visibility inside AI answers – priced like premium media inventory.
2026 midterms
The price of an AI recommendation
What campaigns spent – and were quoted – to show up in and around AI answers
AI-made political ads
$137M+
Spent on 220 AI-generated ads across 34 states ahead of the 2026 midterms.
2024 cycle: under $50,000
AI visibility tracking
$9-$250
Per month: what CampSight charges campaigns to track how LLMs present them – and get tips on influencing it.
GEO strategy pitch
$200K
Visigent’s proposed top success fee for a “category win” inside AI answers.
Sources: Wesleyan Media Project (calls its tally “almost certainly an undercount”); Reuters.
- AI-made ads were the visible line item: in the 2024 cycle, the Wesleyan Media Project identified under $50,000 in spending on AI-generated ads; ahead of the 2026 midterms, that figure was $137M+ across 220 ads in 34 states. Only 27% disclosed AI use, and the project calls its figure “almost certainly an undercount” because AI use is hard to detect.
- Buying AI’s recommendation was a bit quieter: According to Reuters, Austin-based firm Visigent pitched Mike Minogue, a Republican candidate for Massachusetts governor, a GEO strategy that would nudge chatbots to paint him as the preferred choice using tactics such as creating hundreds of AI-friendly policy pages and paid placements on high-authority sites that LLMs already cite, plus Reddit and YouTube content. Visigent’s proposed fee for winning a category: $200,000.
- Both sides of the aisle are buying in: CampSight (an AI visibility tracker for candidates), run by the progressive Run for Something Action Fund, charges campaigns $9 to $250 a month to track how LLMs present them – and how to change it.
Now, note the KPI. Strategist fees were tied to “AI recommendation share” – i.e., how often a chatbot returns the candidate in an answer. Benchmark that against rivals, replace “candidate” with “brand,” and you get one of the core GEO metrics: AI Share of Voice (SoV).
How did AI engines handle election questions?
Responses across models were uneven: ChatGPT and Perplexity were answering election questions directly, and Google was largely withholding AI-generated answers. Critically, where chatbots did answer, they often got basic facts wrong.
2026 midterms
How AI engines handled election questions
ChatGPT and Perplexity answered. Google’s AI mostly stepped back.
| Engine | Answered | Answered vs. declined |
|---|---|---|
| ChatGPT | 20 of 20 | |
| Perplexity | 20 of 20 | |
| Gemini | 19 of 20 | |
| Google AI Mode | 9 of 20 |
AnsweredDeclined
Google AI Overviews
0 of 400
Runs in which an AI Overview appeared – 4 election prompts, 100 runs each, in October.
Sources: EvidentlyAEO (race answers, Nov 1, 2026); Evertune (AI Overviews, October 2026).
In Evertune’s October test (4 factual election prompts, 100 runs each), ChatGPT had the highest share of correct answers on every prompt, Gemini and AI Mode often hedged or deferred to official sites, and AI Overviews never appeared once. Two days prior to the vote, EvidentlyAEO found AI Mode declined to name a favorite in 11 of 20 close races, while ChatGPT and Perplexity answered all 20.
Was Google’s retreat a pattern or a glitch in the Matrix?
Google AI’s behavior was definitely a pattern, and two independent analyses, nine months apart, documented it:
- Around February 2, States United noticed Google started switching election queries to links-only output; non-election queries kept their summaries;
- The day after Election Day, Define Media Group’s VP of AI and Search Strategy, Shahzad Abbas, found no AIOs on any of the 14 most-clicked election queries. Instead, Google served its AP-powered results module and Top Stories.
Pattern, not glitch
Google’s retreat from AI election answers
Four independent checks across nine months found the same behavior
Feb 2, 2026
Links only
Election queries switch to links-only output; non-election queries keep their summaries.
Nov 4, 2026
0 of 14
Most-clicked election queries with an AI Overview; Google showed AP results and Top Stories instead.
Election Day: November 3, 2026.
How accurate were AI answers when they did appear?
In The Washington Post’s review of 3,000+ chatbot voter guides, 48% didn’t name the voter’s congressional district at all and, of those that did, 22% named the wrong one. Some of those errors traced back to small voter-guide sites that appeared to be AI-built themselves.
Where AI did answer
Could chatbots name your congressional district?
The Washington Post’s review of 3,000+ voter guides generated by six chatbots
| Outcome | Share |
|---|---|
| Didn’t name a district at all | 48% |
| Named the wrong district | ~11% |
| Named the right district | ~41% |
Fewer than half got it right. Some errors traced back to small voter-guide sites that appeared to be AI-built themselves.
Source: The Washington Post, Oct 28, 2026 (48% named no district; 22% of the rest named the wrong one). Wrong and right shares derived from those figures.
What’s even more important is what Caucus AI found: 18% of ChatGPT’s September election-related answers cited such sites. This is yet another confirmation that third-party footprint is one of the strongest AI visibility factors – regardless of the source quality.
What can brands and marketers learn from 2026 elections?
Three things brands and their marketers should take away from the 2026 elections: AI recommendations now have a market price, AI engines treat sensitive topics differently, and third-party sources shape what AI says about you.
- AI recommendations are now priced commodities: If entities operating under the heaviest scrutiny online are paying for AI visibility, assume your competitors are at least measuring it.
- Sensitive categories play by different rules: Google withheld generative answers where errors are costliest. Therefore, it’s reasonable to expect similar treatment for health, finance, legal, and other YMYL queries.
- Your third-party footprint is how AI sees you: The fact that ChatGPT repeatedly cited content from anonymous, AI-built sites on a topic this sensitive shows that presence beats quality when authoritative sources don’t respond. Wherever you don’t supply the answer, someone else will, so:
- Monitor what AI engines say about you – and correct the sources they cite;
- Manage your third-party presence the same way you manage your own domains.
Jordan Parkes, the CEO and founder of ZeroClick Labs, has been building and scaling digital marketing strategies since 2012, leveraging performance-driven SEO and data-based digital marketing solutions to guide the growth of hundreds of companies across the U.S. and Europe.
Every buyer’s “Who’s best for this?” gets an AI shortlist
If you’re not on it, you’re losing the race – without even realizing it
Because your analytics don’t show clicks that never happened.
ZeroClick Labs employs a wide range of AI optimization strategies, purposely developed to give brands the highest odds of appearing in ChatGPT’s, AIOs’, and Perplexity’s answers.
The ballot’s already printed.
“Our agency had no idea how to approach AI visibility. ZeroClick only does this one thing so they actually know what works. Worth every penny just to not waste time figuring it out ourselves.” – Jay