8 October 2026 Elena Toniolo
Journal of Experiences
LLMs take information architecture beyond navigation, towards a new way of organizing and making knowledge accessible
For many years, we have associated information architecture with very concrete elements: menus, categories, navigation systems, and taxonomies—structures designed to help people find their way around websites, applications, and digital services.
The advent of Large Language Models seems to challenge this paradigm: if we can ask a question in natural language and get an answer directly, do we still need to design information structures?
At first glance, it might seem not. If a system can interpret a request and return relevant content, then menus, classifications, and navigation paths may seem less essential. But what is changing is primarily what we see of the information architecture, not its function.
The structure doesn’t disappear—it moves behind the interface. And that is precisely why it takes on new significance.
The end of navigation?
For a long time, a key challenge of the digital experience was facilitating access to information—helping people find what they were looking for amidst a growing volume of content.
Information architecture addressed this need by organizing, classifying, and labeling information, creating pathways to guide navigation.
LLMs are changing the way we access information. We do not necessarily need to know in advance where to look or how to phrase a search query; we can start directly with our specific need and express it in our own words. The system interprets the request, identifies relevant information, and uses it to construct a response.
This is a significant shift, yet it does not eliminate the need to organize information. While access becomes more direct for the user, that simplicity masks an underlying need to structure content and relationships so that the right information can be retrieved and used in the correct context.
LLMs do not eliminate structure
Behind a response generated by a language model-based system may be documents, knowledge bases, metadata, and relationships between information that help define the available context.
Even when this organization is not visible to people, it continues to impact the experience.
On a traditional site, faced with an unclear structure, we can attempt alternative routes: going back, exploring another category, using internal search. In a conversational interface, however, much of these references disappear. We ask a question and rely on the system’s ability to identify and relate the necessary information.
Poorly structured content, contradictory information, or unexplicit relationships can therefore impact the relevance and reliability of responses.
Information architecture is not being eliminated by LLMs: more and more of its work is simply taking place out of people’s sight.
From designing paths to designing meanings
Historically, information architecture has helped people find and understand information. With LLMs, this responsibility expands: knowledge must be organized so that it can also be used effectively by systems.
Taxonomies, metadata, and conceptual models help make relationships explicit, distinguish similar information, and reduce ambiguity. It does not mean designing “for machines” in place of people, but considering a new level of mediation between people and information.
The quality of this organization directly impacts the experience: inconsistent or incomplete responses, unrecoverable information, or misinterpreted requests may also depend on how knowledge is structured and connected.
Information architecture thus takes on a connecting role, creating the conditions for systems to use information in the correct context and return it to people in a relevant and understandable way. The structure is not seen, but it influences the experience.
The new questions of information architecture
If the way we access information changes, so do some of the questions that drive design.
The traditional question — where does a person expect to find this information? — is joined by others:
- how should knowledge be organized for a system to retrieve it in the correct context?
- which relationships need to be made explicit?
- what information has priority?
- what concepts can generate ambiguity?
These are not entirely new questions: what is changing is the context in which the information architecture operates and the way in which information is used.
We continue to design for people, also considering the systems that increasingly mediate their access to knowledge.
From websites to knowledge systems
With LLM-based systems, information architecture goes beyond the ability to reach content and helps define how different information is retrieved, linked, and used to construct a response.
The focus thus shifts from the individual page to the system of relationships between contents: where information is located, what it means, what it is connected to, and in what context it is valid become parts of the same system.
A discipline more important than ever
As access to information becomes increasingly immediate, the quality of its organization becomes increasingly crucial. With LLMs, the structure recedes from view, yet it remains fundamental to the quality of the experience.
This is perhaps the most interesting evolution in information architecture: becoming less visible precisely as it becomes more central.
From navigation to knowledge, the way the discipline manifests itself changes, but its task remains the same: organizing information to make it understandable and usable.