Conversational user interface is quickly becoming a new standard of how we interact with digital products and services. The rise of ChatGPT, Claude, and Gemini had a tremendous impact on how people expect digital interfaces to respond to their inquiries. The shift in user expectations impacted all areas of UIs, including one that felt unshakable. I’m talking about the search experience. Users expect modern search to answer questions, not just return links.
In this article, I want to discuss how user expectations lead to this new type of search experience and how you can introduce this new experience to your website today.
Search is dead. Long live AI search
From the early days of the internet, the idea of the search was grounded in this way of interaction:
- Users formulate and submit a search query.
- The system responds to this search query with the list of links that are relevant to the search query.
Before the AI era, this was conventional behavior. This is how the Google Search engine worked, and this is what people expect from all other websites. But OpenAI with ChatGPT disrupted it. ChatGPT proved that people value their time and want to quickly find a solution to their problem.
When it comes to search, people want answers. And they want to use mechanisms that allow them to get answers as quickly as possible (in fact, 73% of visitors abandon digital journeys when they can’t find answers quickly). The format that works best for most users is known as a ChatGPT-style format.
ChatGPT-style format is a conversational format. Users ask questions, and they receive direct answers. They can follow up with additional questions to clarify something or narrow down the search area.
How to add ChatGPT-style AI Search on your website
There are a few ways you can add ChatGPT-style answers on a website. The first option is to use a general-purpose AI assistant model like ChatGPT 5.4 for search.
Although this seems like a reasonable solution, embedding a general-purpose AI assistant introduces several challenges:
- Hallucinations. AI models are prone to hallucination. General AI models may generate answers that aren’t grounded in your content.
- Data privacy concerns. Your content may be sent to external AI providers or used for model training.
- Lack of source transparency. The AI model is a black box, and you have a limited idea of how it gets the answers. There is no such thing as a search log when you use an AI model.
For many organizations, especially those with compliance or data-governance requirements , these challenges are critical. They will likely choose a tool that offers AI answers generated directly from their own website content and provides better privacy and source transparency.