**If you work in technology — or have been remotely in the loop on current events — you've probably heard of, or even used, ChatGPT. (For the record, I did not use ChatGPT to write this blog… I did think about it.) My experimentation with ChatGPT started as a fun party trick. I gave prompts like "Write a Muppets script in the style of The Office" or "Would humans be able to beat the Na'vi from Avatar in a game of basketball?". While I highly enjoyed chats like these, my attention quickly shifted to wondering how this technology would be used in the practical world. Most of us who have played around with ChatGPT have tried using it to answerquestions that we would normally prompt to Google. This behavior has many tech enthusiasts wondering if ChatGPT will threaten Google — and the search industry as a whole. That's a big proposition – but is it true? I decided to run through some use cases and find out for myself. Chat and Search Have Different Strengths and Weaknesses Unsurprisingly, by running a variety of queries, I've found that chat and search each excel in different areas — and they can best be used as compliments to each other. Searches for "encyclopedia-type" knowledge (such as "what is the history of the television?) yielded much better results on ChatGPT than traditional search. ChatGPT excels at delivering thorough, readable, article-style answers to these types of queries.**But you don't have to take my word for it. Funny enough, ChatGPT will tell you itself:
In addition to differences in the type of query they answer best, there are a few other inherent strengths of chat and search that make each technology powerful for specific use cases and situations. Here are just a few: ChatGPT Strengths:
- The user inherently expects chat to be a more conversational, slower experience. A longer time to respond means you can use bigger, more powerful models to generate more detailed responses.
- Chat can take the history of the conversation into account. This means users can ask follow-up questions and dig deep into certain topics. (E.g. chat could help you debug your code.) Search Strengths:
- Search is better for browsing experiences. For example, you would never want to use Chat as the primary interface to your eCommerce site or your store locator. Those are visual browsing experiences, ideal for search.
- Search makes it easier to merchandise/hard-code results based on business logic. This is quite difficult to do in chat — though not impossible.
- It's much cheaper and faster: search is basically free to run. Chat is comparatively more expensive.
- Search will never make anything up. It will only return results based on the information and content that has been indexed in the search engine's database. Which Technology Should Businesses Focus On? The answer here is easy: both. There are massive benefits to each that businesses need to start incorporating into their strategy. Customer experience is critical, and businesses can't afford to lose any opportunities to drive engagement based on poor experiences. With conversational AI (powering chat) businesses are able to provide that streamlined, natural language-based conversation that consumers are looking for. Chat is able to bring context and continue the conversation — unlike search, which is limited from one search to the next. It's worth noting that Chat is just one application of large language models (LLMs), and there are many more. You're even able to ask them to perform arbitrary tasks such as finding a pattern in data and making changes (which is a concept the Yext platform leverages in its AI Data Cleaning transform). With search, on the other hand, businesses can streamline the customer journey and drive engagement. Not only are searches faster, but they can insert valuable calls-to-action, maps, and other rich results that help drive engagement and revenue. How Can Chat and Search Work Together Most Effectively? Generative models are amazing at "understanding" natural language and generating natural responses that seem accurate. However, they also:
- Make stuff up (otherwise known as "hallucinating")
- Can't look up real-time data
- Do not provide a business with control over the answers
- Take a LONG time to train The key to providing an amazing chat or search experience exists with the underlying data. There are different times and places where each can be useful, however, at the end of the day whether you are searching or chatting, you just want the right answer. Sometimes this data is sourced from the entire internet, which is great for answers to the "encyclopedia-type" queries mentioned above. However, for many queries, businesses will want to provide specific data that they are the ultimate authority on, such as their menu items or hours of operation. Enter the knowledge graph. A knowledge graph is a next-generation CMS that is structured with real-world entities and their relationships — and it is incredible at interfacing with natural language experiences. With a knowledge graph, a business can gain control of their content and bolster their digital experiences by serving as an interface between the business and the generative model powering chat, or the algorithms powering search.**
