As the U.S. election approaches, AI search engines face huge challenges, and accurate and reliable information is crucial. The editor of Downcodes will take you to understand the strategies and performance of Perplexity AI, Google and OpenAI in dealing with this information war, as well as the opportunities and risks they each face.
As the U.S. election approaches, AI search engines are facing a major test. At this important time, major technology companies are working hard to provide accurate and reliable information to address concerns about misinformation and AI-generated content "hallucinations."
Amid this competition, Perplexity AI is attracting attention with its new AI Election Center. The platform provides real-time updates, combining data from The Associated Press and a partnership with Democracy Works, a nonprofit voter guidance site. Users simply enter their zip code to get detailed candidate information, ballot measures and answers to related questions. During the test, the reporter entered multiple postal codes in the New Jersey area, and the information displayed was basically accurate.
Unlike Perplexity, Google and OpenAI behave more cautiously when handling election-related queries. Google restricts election-related information on its Gemini website and apps, and users who ask questions receive a mild but ineffective response telling them that they cannot provide information about elections and political figures. OpenAI's ChatGPT Search also takes a conservative approach, directly recommending users to check news sources such as the Associated Press or Reuters for inquiries about the latest polls.
However, Perplexity aims to answer election-related queries as comprehensively as possible. Although the sources of its information are clear, the quality of the sources is sometimes surprising. For example, when asking "what role did battleground states play in the final outcome," Perplexity cited India's Times of India as one of its sources. Reporters found that Perplexity made errors in its polling data in the state, raising questions about the accuracy of its sources. Perplexity responded by saying it would be more careful in evaluating and selecting sources related to U.S. elections.
In addition, Perplexity said it will prioritize fact-checked, non-partisan sources in responding to user inquiries, with the goal of providing at least 7-8 sources per election-related question in order to cross-verify information. However, some information still relies on a single source, such as data on battleground states and electoral votes.
While Perplexity outperforms its competitors in source citation transparency, the accuracy of its real-time coverage remains to be seen. If Perplexity can win the trust of users, it could change the way voters obtain election information and force OpenAI and Google to change their existing caution. However, this process also comes with high risks: any major mistakes could have irreversible consequences for these companies.
All in all, AI plays an increasingly important role in information acquisition, but accuracy and reliability still need to be continuously improved. While pursuing comprehensive information, major companies should also pay attention to the accuracy and reliability of the source of information so that they can truly provide value to users and avoid negative impacts. The editor of Downcodes will continue to pay attention to the developments in this field.