Ask instead of search

by Stefan Grill, Head of Innovation at 3pc

Googling and then trawling through endless lists with thousands of search results? Far too exhausting - when I can simply type a question into ChatGPT and get a well-formulated answer in seconds ...

The memory of how ChatGPT amazed the public in November 2022 is still quite fresh, with a few "teething problems" at the beginning. In the meantime, it has become so commonplace and mature that you sometimes forget you're talking to a machine in natural language. Thanks to their constantly growing "knowledge", chatbots are also getting better at recognising the intention behind a question, which is another major advantage over sometimes frustrating internet searches.

Changing role of websites

Chatbots are now taking over far more than just chats; the path is moving towards AI agents that decide and act independently on our behalf. Almost unnoticed, we are increasingly handing over control to ChatGPT and its ilk.

We can decide for ourselves whether we want this. The situation is different for website operators: This is because actual (human) visits to the online presence decrease to the extent that AI intervenes as an explanatory or acting medium. The purpose of websites is changing significantly and they are becoming secondary sources of information. As only a few people check the sources of the bots themselves, traffic decreases - at the same time, the information sovereignty over the knowledge provided is jeopardised.

Incorporating new usage habits

What do these developments mean for website operators? It means adapting to the new user habits and creating even more attractive access to their own content. Access points that make it easier for visitors to browse through the content themselves: "Answer machines" can also be docked onto your own website, needs-orientated AI agents that act as interactive user interfaces to support the search for relevant information.

Instead of relying on often abbreviated answers from general chatbots, visitors to the website can then search through content in natural language and receive answers based on reliable information. An offer that invites visitors to delve deeper into the content presented.

Own side chatbot thanks to RAG system

From a technical point of view, so-called RAG systems are state-of-the-art for such searches in natural language on your own website. RAG stands for Retrieval Augmented Generation. Translated, this means that the system first searches for content that matches the user input (retrieval). The results are sent to an AI model, which incorporates the content into its "knowledge" and generates a textual summary (augmented generation).

This approach opens up new possibilities that go beyond the strategies of leading AI companies. Their offerings aim to ensure that their users are content with what the friendly AI chatbot outputs.
Websites with their detailed content form an antithesis to this and are still the core of the World Wide Web: an open and democratic internet in principle.
RAG on websites imports the new usage habits directly into online presences and can thus help to preserve the openness of an internet without pre-filtered knowledge.

Digital sovereignty

The increasing number of freely available AI models also has a positive effect on the development of such RAG systems; this reduces the dependency on known quantities such as GPT4 from Open.ai (ChatGPT).

At 3pc, we plan and develop solutions in such a way that they work both connected to large commercial AI models and with freely available ones. You can also switch - for example, to follow changing legal or ethical compliance guidelines or to save costs. In this way, the website chatbot via RAG also promotes your own digital sovereignty.

RAG data for the content strategy

Instead of leaving the selection of sources to large AI chatbots - which can certainly lead to incorrect information or even errors by the AI - operators of their own website RAGs retain control over the information and data they offer.
And they can see which search queries users have made - an excellent basis for further improving both their own AI agent and the website content.

In the future, a RAG system could even be used to connect additional sources of information such as publications, collections or knowledge databases. Instead of putting a lot of effort into bringing these onto the website, they could be made available in a user-friendly way via the AI agent on the website.

How a web of AI agents will develop in the coming years remains to be seen. However, we can already assume that the topic of "search engine optimisation" will develop rapidly against this backdrop.

We at 3pc are also working on this and can provide guidance [LINK GEO-Audit]: Your own content must be relevant where users are, with the next generation of search engines. So that human visitors arrive at your website - where they can then optimally satisfy their need for information.