Editorial illustration for OpenAI Details 2026 Election Stance as Chatbots Rely on Problematic Data
OpenAI's 2026 Election Stance: Chatbots' Data Problem
Voters heading to the polls today have company they probably shouldn't trust. Pew Research found that roughly half of US adults have used an AI chatbot at some point, and 8 percent turned to one specifically to research candidates and policies ahead of the midterms. The problem: these tools are getting basic election facts wrong, often with total confidence.
Forum AI's research puts a number on it. Nearly 40 percent of election-related answers from major chatbots, including ChatGPT, Claude, and Gemini, contained at least one factual error. Google's AI Mode reportedly refused to give results from the 2020 election outright.
Gemini told users the Pennsylvania mail-in ballot deadline was November 3, nine days after the actual cutoff of October 27. In August, Claude mixed up North Carolina Congressman Don Davis with a former school superintendent who happens to share his name.
Google and Anthropic didn't respond to requests for comment on their chatbots' errors. OpenAI pointed WIRED to a blog post laying out its election policy for 2026. The companies building these tools market them as neutral, fact-based alternatives to partisan noise. The data behind them tells a messier story.
According to the Forum AI research, almost 40 percent of election answers given by all the mainstream chatbots contained at least one factual error. In August, when Forum AI’s researchers asked why someone should vote for North Carolina congressman Don Davis, Claude gave a response that confused the lawmaker with a “former school superintendent” of the same name.
Why this matters
For anyone building on top of these models, the NewsGuard finding should sting more than OpenAI's election blog post does. Nearly half of citations on contested political queries tracing back to "pink slime" sites isn't a content gap you patch with a disclaimer. It's a sourcing problem baked into how these systems retrieve and rank information, and it's happening at the exact moment 8 percent of US adults are using chatbots to decide how to vote.
OpenAI pointing reporters to a policy page rather than addressing the sourcing issue directly tells us where the industry's priorities sit right now: PR framing over retrieval accountability. If you're a developer shipping a RAG pipeline or a founder building an assistant that touches current events, this is your warning that "it cited a source" means nothing if nobody's auditing what kind of source. Researchers should be asking NewsGuard-style questions of their own systems before launch, not after a Pew survey catches the fallout.
The midterms are a stress test. The real question is whether anyone fixes the pipeline before the next one.
Related Reading
Browse all OpenAI news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google to End Free Access to Gemini AI for Non-Subscribers
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic OpenAI AGI chief takes leave as company buys viral talk show TBPN today
Business & Startups
Related Topic OpenAI's GPT Transcribe Cuts Error Rate to 3.31%, Improving on GPT-4o
Open Source
From Archives Apple's first robot proves harder than expected; AI robots still far from useful
From Archives NVIDIA Cosmos Platform Aims to Simplify World Model Development
From Archives OpenAI raises bug bounty payouts to USD 172,000, matching Google rates
Business & Startups
From Archives OpenAI pitches data center power plants as protectors amid US opposition
Business & Startups
Common Questions Answered
What percentage of election-related answers from major chatbots contain factual errors according to Forum AI research?
According to Forum AI's research, nearly 40 percent of election answers given by mainstream chatbots including ChatGPT and Claude contained at least one factual error. This high error rate is particularly concerning given that 8 percent of US adults are using chatbots specifically to research candidates and policies before voting.
Can you provide an example of how chatbots confused election candidates?
In August, when Forum AI researchers asked Claude why someone should vote for North Carolina congressman Don Davis, the chatbot confused the lawmaker with a former school superintendent who shared the same name. This demonstrates how chatbots can confidently provide incorrect information about specific political figures and their backgrounds.
What is the sourcing problem with how chatbots handle contested political queries?
NewsGuard found that nearly half of citations on contested political queries trace back to 'pink slime' sites, which represents a fundamental sourcing problem baked into how these systems retrieve and rank information. This issue cannot be solved with simple disclaimers, as it reflects deeper problems in the training data and information retrieval mechanisms of these AI systems.
How many US adults have used AI chatbots for election research according to Pew Research?
According to Pew Research, roughly half of US adults have used an AI chatbot at some point, with 8 percent specifically turning to one to research candidates and policies ahead of the midterms. This significant usage rate makes the accuracy problems with election-related answers particularly problematic for democratic decision-making.
Further Reading
- AI chatbots are becoming political advisers. Their answers can mislead voters - WFAE
- OpenAI starts election ‘war room’ to monitor its products - Washington Examiner
- Few Americans Say They Are Asking AI Chatbots for 2026 Election Information - Pew Research Center
- OpenAI launching election 'war room' - Yahoo News
