Tech Investor Kyle Corbitt Highlights AI Chatbots' "Inadequate" Conversation Search, Urges "Agentic Search" Adoption

San Francisco – Tech founder and investor Kyle Corbitt has publicly called out leading artificial intelligence platforms ChatGPT and Gemini, criticizing their current "keyword-based search for finding the correct prior thread" as insufficient. Corbitt, who engages in "20+ conversations a day" with these AI tools, stated via tweet, "> @ChatGPTapp @GeminiApp whichever one of you gets proper agentic search working over my past conversations first wins my business. I do 20+ conversations a day and your keyword-based search for finding the correct prior thread sucks." His remarks underscore a growing demand among power users for more sophisticated conversational memory and retrieval systems.

The critique highlights a significant user experience challenge for individuals heavily reliant on AI chatbots. Users frequently report frustration with the basic search functionalities offered by both ChatGPT and Gemini, often finding it difficult to locate specific information within extensive chat histories. This limitation often forces users to manually scroll through conversations or resort to third-party extensions to manage their interactions effectively.

"Agentic search" represents a more advanced AI capability where systems autonomously pursue complex goals, make decisions, and execute multi-step processes with minimal human oversight. In the context of conversational AI, this would involve large language models (LLMs) intelligently breaking down complex user queries, leveraging comprehensive chat history for deep context, and synthesizing relevant information from various sources to provide precise and proactive results. Microsoft's Azure AI Search is an example of a system developing such agentic retrieval capabilities.

While Google's Gemini platform has introduced features allowing it to reference past chats for more helpful responses and offers users control over their data, its direct search functionality within chat history remains rudimentary. Similarly, OpenAI's ChatGPT provides a basic search, but user feedback consistently points to its inadequacy for extensive archives, despite recent developments like the "Pulse" feature for personalized daily summaries. The lack of robust, context-aware search stands in contrast to the advanced capabilities these LLMs demonstrate in other areas.

Corbitt's background as a co-founder of companies like Socialcam and Loopt, and a former Partner at Y Combinator, lends considerable weight to his public commentary. His challenge to OpenAI and Google emphasizes a critical need in the competitive AI landscape: the development of intelligent, agentic search features that can transform how users interact with and leverage their vast AI-generated conversational data. This demand is likely to spur further innovation as platforms vie for user loyalty through enhanced utility.