Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Get the latest local stories in your inbox.
Millions of Americans are turning to AI-powered chatbots this midterm season to figure out where and how to vote, and behind many of those answers sits a nonprofit most voters have never heard of, operating out of a small office in Brooklyn. Democracy Works has quietly become a primary source of election information for some of the world’s largest technology companies, even as the group says it’s struggling with funding.
According to NBC News, the small nonprofit based in Brooklyn now partners with Google, Anthropic and OpenAI, supplying the data those companies’ chatbots use to answer questions about registration, polling places and deadlines. Democracy Works describes its own role in blunt terms: its Elections API, the group says, “provides the comprehensive election data that AI platforms deliver at scale,” per Democracy Works. The organization says that dataset covers every upcoming federal, state, county, municipal, sub-municipal and school board election in jurisdictions with more than 5,000 people, and that it continuously monitors changes while working directly with state and local election officials.
OpenAI has detailed its own side of the arrangement. The company says its Democracy Works integration is now live across the United States and in multiple languages, bringing registration deadlines, voting locations and other requirements directly into ChatGPT, according to OpenAI. OpenAI adds that Democracy Works compiles that information with state and local election officials and monitors it for changes throughout the election period, and that ChatGPT’s responses include links to official election sources so users can verify details themselves.
Despite its outsized role feeding data to some of the most heavily used software on the planet, Democracy Works says it is struggling with funding, per the nonprofit’s own account relayed by NBC News. The group has described operating in what it calls a sea of 10,000 U.S. election jurisdictions, a fragmented landscape the organization says its research team works to track and verify as changes occur. This year, the same report notes, Democracy Works released an updated AI Manual offering generative AI systems practical suggestions for responsibly integrating election data.
It isn’t the nonprofit’s first brush with major platforms. Democracy Works says it also struck successful 2024 integrations with Meta, TikTok and Perplexity AI, extending its reach well beyond the three partners named in the current reporting.
The surge in chatbot-assisted voting research is real but modest in scale. A Pew Research Center survey of 5,273 U.S. adults conducted October 12 to 18 found that 8% said they were using AI chatbots for election-related information this year, according to Pew Research Center. Among those users, 76% said they turned to chatbots to learn where candidates stand on issues and 71% used them to check information encountered elsewhere, while smaller shares, 40% each, sought vote-choice suggestions or logistical details like registration and polling locations. Pew also found a generational split, with 11% of adults under 50 using chatbots for election information compared with just 4% of those 50 and older.
A separate September survey cited by The Atlantic found an even smaller share, just 4% of respondents, had consulted a chatbot for election information in the prior month, while nearly 40% said they’d gotten such information from Google’s AI-search summaries instead, according to The Atlantic.
Even as adoption grows, testing this fall has repeatedly found gaps in reliability. Forum AI’s research, detailed in the same Atlantic report, found that about 40% of chatbot answers to election-related questions contained at least one factual inaccuracy, and that roughly one in three responses cited an unreliable source such as posts on Reddit or Quora.
NPR’s review of October data found 13.9% of chatbot answers to voting-logistics questions were outdated or inaccurate, with errors in 5.9% of answers serious enough to potentially mislead a voter and 5.1% of answers overall containing a clear factual error, according to NPR. Accuracy diverged sharply by platform in that testing: researchers found Claude gave wrong or outdated voting information about one in three times, Copilot about one in four, and ChatGPT about one in ten. The same testing found that every chatbot except Google’s Gemini offered lists of candidates to consider based on a user’s stated political leanings, and that all of the platforms commonly cited prediction-market sites like Polymarket and Kalshi when asked who was winning a race.
An Institute for Strategic Dialogue study put the overall error rate even higher, finding 29% of chatbot responses to election questions were incomplete, inaccurate or outdated, as reported by The Hill. That same study found accuracy dropped by 16 percentage points when questions were posed in Spanish, and ranked OpenAI’s GPT-5.5 as most reliable at 89% specific, complete and correct answers, followed by Google’s Gemini 3.5 Flash at 84%.
Not every assessment points in the same direction. States United reported that factual errors in the AI platforms it tested fell from roughly 7% to 8% in its first round of testing last fall to zero in its second round this year, according to States United. Even so, the group found that platforms pointed users to their state’s official election website less than half the time overall, and almost never for certain questions. One exception stood out: when asked for the nearest polling location, ChatGPT mentioned the state’s election website 88% of the time, while Google AI did so 100% of the time in incognito mode, per the same States United analysis.
Those mixed findings underscore the stakes riding on a single small nonprofit’s data pipeline. With Democracy Works warning that its own funding is uncertain, the infrastructure quietly underpinning how millions of Americans get voting answers from ChatGPT, Gemini and Claude may be more fragile than the scale of its reach suggests.
Science, Tech & Medicine in …
About 12 hours ago
217 Affordable Apartments Open for Lottery at Former Greenpoint Hospital Site
About 14 hours ago
Chelsea Hotel Site Finally Flattened After 12-Year Fight Over Mixed-Use Tower
About 14 hours ago
Yonkers Dashcam Captures Shoulder-Riding Altima T-Boning Driver Mid-Turn
About 15 hours ago
Police Seek Suspect After Knifepoint Robbery on East 86th Street
Hoodline – In-depth reporting about your home area.
Follow Us
Connect
But wait, there’s more…
Cities
Please enable JavaScript to select your regional newsletters and subscribe.
Choose one or more regions. Get the latest local stories when there is news to share.
Select the local newsletters you’d like to receive, then subscribe.
0 regions selected
No matching regions. Try another search.
Free. Unsubscribe anytime. Daily when new stories are available.
Open the standalone signup form
Subscribe to Hoodline newsletters