17 January 2026 by Mondo2.0

The Hyper-Mediation of Knowledge: When LLMs Replace Search Engines

For over a decade, Google has been the go-to source for any online search or in-depth analysis in Europe. To find a news item, you simply typed it into Google and selected, in most cases instinctively, one of the first three results. During the same period, for public institutions and private companies, organic SEO was the key tool for dissemination and visibility.

I need to engage Google, share my content with them; if it’s well-described, original, and unique, I have a chance of appearing in their search results.

Properly optimizing your content therefore meant being visible on the Internet, reaching new audiences—not just those who already know you—attracting visitors, and building interest and a community around your publications.

Statistics show that organic visibility expanded, and still expands today, the Internet reach of institutions and companies that would otherwise have remained confined to a local or specialized audience.

But what is organic SEO (Search Engine Optimization)?
It involves optimizing web content without paying for advertising.

For cultural institutions, in my professional field, this meant structuring and describing websites, events, exhibitions, digital catalogues, artwork profiles and archive documents in the best possible way, so that they could be suggested by Google and, consequently, selected by those seeking cultural information.

Today, after more than a decade of being “Google-centric,” this balance between content offerings and listings to capture the attention of Internet users is eroding.

The advent of Large Language Models and their applications in search engines is redefining the way knowledge is mediated online.

There’s not just Google to ask: there’s ChatGPT, Claude, Gemini, … in addition to dozens of specialized conversational tools for industry professionals, teachers, and students.

Furthermore, these tools no longer simply connect information supply and demand by offering a “simple list of opportunities”: they summarize content, directly respond to users’ questions, and in the process intercept visitors who previously accessed the original sources, replacing open-access search tools based on search engines.

The social change affects all information searches: queries in which users seek specific knowledge on the Internet.

It’s precisely these searches that are intercepted by LLMs, providing immediate answers without requiring the user to visit the source site. Rather, they provide a dialogic opportunity to ask further questions and gain progressive insights, actively guiding the path of discovery.

The main consequences of this change:

• There is a plural offering of search tools: the traditional search engine is no longer the only gateway to request information and to select what to explore further;
• LLM training does not follow a linear logic; the process of learning and providing information is subject to many more variables and alternatives, including the contents of the dialogue between the user and the conversational agent, which are previously unpredictable.
• Public and private institutions produce quality content that feeds the training of AI models, but they no longer receive the access necessary to financially support that production, because the conversational agent often responds directly.

Will websites disappear in 10 years? Will they be completely replaced by conversational agents?

From a sociological perspective, we are moving from a single model, in which Google as a search engine represented an obligatory step, a bridge to sources, to one of stratified mediation in which LLMs intervene as new gatekeepers.

Following this important change, how is the traceability of knowledge ensured?

To be fair, this problem also existed with search engines. Google (at least in Europe) served as a necessary filter, an intermediary used by everyone, or almost everyone, to access information. We’re witnessing a major shift in search methods and options, but at the same time, we’ve been navigating a channelized process of searching and selecting information for a long time now.

The new model changes the way we interact, opening up conversation and dialogic mediation with new players; it’s no coincidence that Google itself immediately revised its interaction method, offering suggestions in natural language before the results list.

If the list of results automatically and instinctively directed us to the first occurrences on the list (often placed for a fee), the dialogue with conversational AI agents can facilitate knowledge but, at the same time, orient our cognitive path in ways that are not always transparent.

On the one hand, conversation allows for deeper exploration of complex topics through follow-up questions, adapting to the user’s level of understanding and suggesting unexpected connections between topics. On the other hand, this same mechanism can lead the user to simplified answers, reduce exposure to different sources, or favor certain interpretations over others, subtly shaping the way we think and the conclusions we reach.

And, as mentioned previously, organic SEO strategies will certainly be less effective when faced with ever-changing conversations and extremely complex learning methods.

Google chose the occurrences to propose at the beginning of the list (in particular the first three, which were by far the most selected), the AI ​​​​chooses for us which information to include, which to exclude and how to present it.

What’s at stake isn’t just gaining bottom-up visibility by capturing the attention of this new generation of conversational agents. What’s at stake isn’t just the flow of visitors and organic access to information from sources, but the survival of a diverse information ecosystem.

Can AI services outperform Google?

Hopefully yes, the arduous sentence for posterity.

Mondoduepuntozero

Read more: “New Professions: the Prompt Manager“