AI chatbots can make election information easy to ask for, but their answers may be wrong, outdated, or drawn from sources that appear to be mainstream news while having political or ideological funding. Research described in the source article found problems across several widely used chatbots, including errors about election results, voting deadlines, and candidates.
Fluent answers can still contain errors
A chatbot may present a response in a confident, conversational way. That presentation does not establish that the information is current or correct. The source article reports that almost 40 percent of election answers given by mainstream chatbots contained at least one factual error, according to Forum AI research.
The examples show how consequential a detail can be. Google’s AI Mode refused to provide the results of the 2020 elections. Gemini gave November 3 as the deadline to request a mail-in ballot in Pennsylvania, while the article says the actual deadline was October 27. In another example, Claude confused North Carolina congressman Don Davis with a “former school superintendent” of the same name.
These are different kinds of mistakes: an answer may be missing, a date may be wrong, or information about one person may be mixed up with someone else. For a voter trying to compare candidates or understand a policy, any of those errors can make an otherwise useful-looking answer misleading.
The sources behind a response matter
Chatbots depend on information available online. The source article says NewsGuard found that mainstream chatbots cited so-called “pink slime” websites nearly 50 percent of the time when asked about candidates and policies related to the midterms. These sites present themselves as mainstream news outlets but are funded by political or ideological groups and individuals.
A source that looks like a news site can therefore require more scrutiny than its appearance suggests. If a chatbot repeats a claim from a partisan source without making that context clear, a reader may not know who is behind the information or what perspective it represents. A polished summary does not remove the need to check where its claims came from.
NewsGuard’s results also varied by chatbot. The article identifies OpenAI’s ChatGPT as the worst offender in that research, citing “pink slime” sites in more than 70 percent of searches. xAI’s Grok performed best, citing them in 30 percent of cases. Those findings describe the research reported in the source; they do not guarantee that any one answer from a particular chatbot is accurate.
What voters should take from the findings
The source reports that 8 percent of recently surveyed US adults said they had used AI chatbots ahead of the midterms to find information on candidates and policies. It also says around 50 percent of recently surveyed US adults had used an AI chatbot at one time. Those figures make the quality of election answers a practical concern: people may encounter them while trying to learn about the choices on a ballot.
Nina Jankowicz, executive director of the American Sunlight Project and a former disinformation czar during the Biden administration, warned that chatbots can make things up when they do not know an answer. Her point, as presented in the article, is that these systems rely on information from sources that may not be good and can produce a prediction instead of a dependable fact.
For voters, the sensible response is to treat a chatbot answer as a starting point. Check election dates and candidate details against sources that clearly identify where their information comes from. For claims about policy or a candidate’s record, look for supporting material rather than relying on a chatbot’s summary alone.
Convenience does not replace verification
Chatbots can gather information into a quick reply, but the examples in the article show that speed and readability are not the same as reliability. An answer can combine old information, a mistaken identity, or a questionable source without making the problem obvious to the person reading it.
That makes verification part of using AI for election research. When the question concerns who is running, what a policy says, or when a voting deadline falls, confirm the details before acting on them. A chatbot can help frame a question, but voters still need to judge the evidence behind its answer.