Perplexity, an AI-powered search engine, can feel hallucinatory at times. But the company wants to show that it’s trustworthy enough to be used to track election results.
Today Perplexity announced a hub dedicated to U.S. general election information. Filled with data from the Associated Press and Democracy Works, the company described it on its blog as “an entry point to understanding key issues.”
“Starting Tuesday, we will be providing real-time updates on the election using data from The Associated Press,” Perplexity said. “This allows us to stay informed about presidential, Senate and House races at the state and national levels,” he wrote. “We are grateful to Democracy Works for giving us access to their Elections API to power this experience.”
Perplexity did not immediately respond to TechCrunch’s request for information about whether it compensates AP and Democracy Works for the data.
Perplexity’s Election Hub not only answers election-related questions like voting requirements and voting times, but also provides AI summary analysis of ballot measures, candidates, policy positions and endorsements. Aside from the summary, the hub is essentially a wrapper around API data from The Associated Press and Democracy Works. However, it is notable that Perplexity’s competitors have been reluctant to release similar features for fear of AI-generated misinformation.
In the recently launched ChatGPT search experience, OpenAI is directing users inquiring about election results to the Associated Press and Reuters. Anthropic’s Claude chatbot doesn’t answer questions about the election results, and neither does Google’s Gemini.
Considering AI’s poor track record in this space, we can only hope that Perplexity’s hub is as accurate as the company claims.
In a July study, the Center for Democracy and Technology found that more than a third of the answers generated by AI chatbots, including Claude and Gemini, in response to 77 different election-related questions were incorrect. I found it to contain information. Other studies have shown that major chatbots perform more poorly when people with accessibility issues and people for whom English is not their primary language ask questions about the election.
TechCrunch has a newsletter focused on AI! Sign up here You can receive it in your inbox every Wednesday.
