OpenAI’s highly anticipated search product, ChatGPT Search, was released last week and aims to take on Google as its main competitor. While the feature shows promise in answering long, written-out research questions, it struggles with everyday searches that make up the bulk of users’ internet activities.
During a test run, I found ChatGPT Search to be unreliable for short, navigational queries, often producing incorrect or irrelevant results. These types of searches are crucial for users like me who make quick, keyword-based searches every day. Google’s 25-year-old product excels in this area, and OpenAI will need to improve its search capabilities significantly to catch up.
The issue lies partly with ChatGPT Search relying on Microsoft Bing, which is widely regarded as inferior to Google. Large language models like ChatGPT may struggle with short prompts due to their reliance on statistical patterns that require fully written-out questions.
Perplexity’s AI search tool, another contender in the market, faces similar challenges. Its CEO noted that users typically ask longer questions than those used for web navigation, which is a critical aspect of Google Search.
While ChatGPT Search is good at surfacing information buried in traditional searches, it cannot replace Google Search for its core function: web navigation. To truly challenge Google, OpenAI must focus on developing a better interface for everyday searches.
As the search landscape continues to evolve, it’s clear that AI products like ChatGPT Search and Perplexity have their strengths but also need significant improvement in areas where they currently fall short. Until then, Google remains the mainstay for many users.
Source: https://techcrunch.com/2024/11/04/chatgpt-search-is-not-openais-google-killer-yet