The influence of artificial intelligence (AI) on how we navigate the web is poised to become increasingly significant, particularly with the emergence of OpenAI’s SearchGPT. This AI-powered search tool aims to provide direct answers to user queries, eliminating the need to sift through pages of optimized search results. While this concept may seem familiar, Google has already experimented with a similar approach using its Gemini AI model, which powered the “AI Overviews” tool. Like SearchGPT, this tool aimed to summarize relevant information from the web, providing users with concise answers to their queries. However, Google’s initial attempt fell short, facing criticism due to inaccuracies and questionable results. In some instances, Google’s AI offered nonsensical advice, such as adding glue to pizza sauce or suggesting washing clothes with chlorine. This raised concerns about the AI’s ability to differentiate reliable information from misinformation or satire.
SearchGPT, leveraging the more mature ChatGPT AI model, holds the potential to overcome these limitations. However, it remains in its prototype stage, and its real-world performance is yet to be determined. Despite these uncertainties, the emergence of SearchGPT raises critical questions about the future of AI in web search. Could AI ultimately replace traditional search engines, or will the accuracy of AI-powered search remain a challenge?
Nell Watson, an AI researcher at the IEEE, highlights the inconsistencies inherent in current AI models. These models, she explains, often lack a cohesive thinking process and can generate illogical or even bizarre results. However, ongoing research focuses on developing more effective AI models that exhibit robust reasoning capabilities. These models can be integrated with logical programming languages, such as Prolog, to enhance their reasoning abilities and ensure greater reliability. Watson further emphasizes the importance of “agentic” AI models—models that can understand a situation, formulate plans, and act independently in response. Without such a framework, she warns, AI systems will struggle to deliver accurate and trustworthy information.
The concerns surrounding AI’s role in search stem from the potential lack of a robust framework that can guarantee accuracy and trustworthiness. The desire to quickly capitalize on the burgeoning interest in AI may have led to premature deployment of tools without adequate testing. Watson underscores this point, stating that some AI features were rolled out too early. Beyond the AI models themselves, the shortcomings of current web search engines also contribute to the challenges faced by AI search. Watson points to issues like the prioritization of paid search results and the inherent biases in AI infrastructure that suppress “undesirable” content. She emphasizes the need to prioritize user experience and address these issues to optimize search engine performance.
The future of AI in search appears to lie in the development of agentic models—autonomous AI systems capable of conducting sophisticated searches and analyses. These models hold the potential to provide answers to questions that users may not even know how to ask. However, as Watson stresses, it is crucial for these AI systems to understand human values, boundaries, and contexts to ensure responsible and ethical operation.
While the challenges presented by AI search are real, there’s significant potential for AI to revolutionize how we access and consume information. Watson envisions a future where AI agents act as our ambassadors, proactively seeking out products, services, and experiences that align with our interests and preferences. In this scenario, traditional search would become obsolete, as AI agents would seamlessly deliver personalized recommendations without requiring explicit user searches.
This optimistic vision comes with a caveat—the need to maintain user trust. Watson cautions against overwhelming users with AI-generated content, emphasizing the importance of respecting user boundaries and avoiding excessive promotional efforts. The future of AI in search will depend on building trust and ensuring that AI systems remain in service of users, not the other way around.
