> ## Content Index > Fetch the complete content index at: https://rabagas.ghost.io/llms.txt > Use this file to discover other available public pages before exploring further. # The Audience We Never Invited - URL: https://rabagas.ghost.io/the-audience-we-never-invited/ - Published: 2026-09-18T13:15:45.000Z - Updated: 2026-09-18T13:15:45.000Z - Description: We built the internet for people. Now machines are reading it, writing to it, and increasingly reading what other machines have written. The audience we never invited is already here. - Author: JTV - Tags: Artificial Intelligence, Internet, Technology, Digital Culture **There is a strange assumption hidden inside the internet, one so ordinary that we rarely notice it anymore: when we publish something, we imagine that someone reads it.** *By Ulrikk K. Reen* Usually, we imagine that someone is a person. Perhaps a bored person sitting on a train, a curious person who followed a link from somewhere else, somebody looking for a particular piece of information, or somebody who has made the regrettable decision to read an argument in a comment section at two in the morning. The reader may be interested, irritated, confused or completely indifferent. They may read the entire thing, skim it, misunderstand it or close the page after twelve seconds. That has been the basic social arrangement of the web for a long time. People make things, and other people encounter them. It was never entirely true, of course. The internet has always had machines crawling through it. Search engines have been indexing pages for decades. Bots have collected information, checked links, measured traffic and performed all sorts of obscure jobs that nobody thinks about until one of them stops working. But there was a useful distinction. Machines helped us navigate the human internet. Increasingly, machines are becoming inhabitants of it. And that distinction is worth paying attention to. ## THE INVISIBLE READER There is something peculiar about being read by a machine. A human reader leaves traces that we recognise as human. They click something. They send it to a friend. They write a comment. They subscribe. They come back six months later because they remember reading something interesting and have completely forgotten where. A machine does not necessarily do any of these things. It can arrive at a page, extract its contents, identify names and subjects, compare statements with information from thousands of other pages, store representations of what it found and move on without leaving anything that looks like a reader behind. The author may never know that the page was read. This is not particularly sinister. In many cases it is precisely what the system was designed to do. Search engines need to read pages if they are going to find them. Archives need to copy things if they are going to preserve them. Information systems need to process information if they are going to make use of it. But there is a curious psychological difference between knowing that something is public and knowing that it is being consumed. When a newspaper publishes an article, it knows that thousands of people might read it. When somebody posts an obscure essay on a personal website, they may imagine that nobody reads it at all. The machine does not care about the size of the audience. If the page is accessible, it can be read. That means the internet contains an enormous population of readers that most of us never see. We are still writing as if the room is empty, while there are machines quietly taking notes. ## THE INTERNET WAS BUILT FOR PEOPLE The original architecture of the web was not designed around the idea that machines would eventually become one of its principal audiences. Again, this requires some qualification. The web has always depended on machines. Search engines, databases and automated indexing are not recent inventions. The machine has been part of the system from the beginning. But the machine traditionally occupied a supporting role. You wrote the article. The search engine helped someone find it. You uploaded the photograph. The server delivered it. You published the document. A database stored it. The machine was the infrastructure around the human activity. That relationship becomes less straightforward when the machine is capable of consuming the information and producing something new from it. An AI system can read an article, summarise its argument, compare it with other articles, extract its claims, rewrite those claims in another style and incorporate the resulting information into another process. It does not simply point a human towards the shelf. The librarian has learned to write. This creates a rather peculiar situation. We have spent several decades building an enormous information system for humans, and we are now filling it with systems that are exceptionally good at reading the information system itself. The web is becoming both the library and the raw material. ## WHEN MACHINES READ MACHINES This becomes more interesting when the machines stop merely consuming human material. Imagine a relatively innocent chain. A person writes an article about a subject. An AI system reads the article and produces a summary. Another system encounters the summary and uses it as part of an analysis. The analysis is published somewhere online. A search engine indexes it. Another AI system later retrieves the analysis in response to a question. By the time a human encounters the final answer, the information may have travelled through several layers of machine processing. Nothing necessarily went wrong. Nobody necessarily lied. Nobody necessarily intended to distort anything. And yet the original human statement may now be several steps removed from the thing being presented to the reader. This is one of the more interesting consequences of machines becoming both readers and writers. The web used to have a relatively simple relationship between authors and audiences. It was possible, at least in principle, to trace a statement back to the person who made it. That becomes more complicated when machines continuously transform material and publish the transformations back into the same environment from which future machines will retrieve information. At some point, the question is no longer simply, *Who wrote this?* We also have to ask, *How many times has this information been through a machine before I encountered it?* And perhaps the stranger question is: *Who was it written for?* ## WE ARE STILL WRITING AS IF THEY AREN'T THERE Human writing contains enormous amounts of information that is not explicitly written down. We rely on context. We expect the reader to recognise a reference. We use irony without announcing that something is ironic. We leave things deliberately ambiguous. We assume that a reader understands that a particular sentence should not be interpreted literally. Human communication is full of shortcuts because humans are very good at filling in the missing pieces. Machines are increasingly good at this too, but not necessarily in the same way. A machine can take a sentence and compare it against millions of other sentences. It can identify recurring names, subjects and relationships. It can discover that two people mentioned in completely different contexts appear together unusually often. It can notice patterns that would be invisible to an individual reader. This creates a peculiar inversion. We often think of machines as being bad at understanding the things that humans understand instinctively. At the same time, machines can notice relationships between pieces of information that no individual human would ever have the time, patience or inclination to examine. A human reader might remember an article. A machine can remember the relationship between that article and thousands of others. The difference is enormous. And it means that publishing something online increasingly involves an audience that is not merely larger than we imagine, but fundamentally different from the audience we imagine. Perhaps it is useful to think of this as a second internet. Not a separate network, and certainly not a place where machines have their own little websites hidden behind ours. The physical infrastructure is mostly the same. The pages are the same pages. The servers are the same servers. What is different is the layer of activity taking place on top of it. The human internet is made of conversations, photographs, articles, advertisements, arguments, jokes, instructions, diaries, propaganda, recipes and the occasional person explaining at considerable length why everybody else is using the wrong kind of screwdriver. The machine layer is made of indexes, embeddings, databases, extracted facts, metadata, classifications, relationships and representations of all those human things. The two layers overlap. Increasingly, they feed each other. A human creates something. A machine reads it. The machine transforms it. Another machine reads the transformation. A human eventually encounters the result and responds to it. Another machine reads that response. The straight line from author to reader becomes a loop. This may turn out to be one of the defining characteristics of the next phase of the web. Not that machines have replaced human communication, but that human communication is increasingly taking place inside systems that also communicate with machines. ## THE QUESTION OF PRIVACY We have historically thought about privacy in terms of people. Did somebody read my message? Did somebody see my photograph? Did somebody download my document? Did somebody copy my work? These are still important questions. But there is another question becoming increasingly relevant: Did a machine ingest it? That is not quite the same thing. A person reading something is an event. A machine processing something can become infrastructure. Once information has been processed, it can potentially be classified, indexed, correlated, transformed and incorporated into another system. The important question is not always whether somebody deliberately stole the information. Sometimes the more mundane question is whether the system was designed to consume whatever information it could access. That distinction matters because the social meaning of publishing has changed without the word itself changing. When we say that something is public, we tend to imagine that we have made it available to other people. But a public piece of information is increasingly available to systems that can do things with it that no individual reader could realistically do. A person might read ten thousand articles. A machine can process ten million. The difference is not merely quantitative. It changes what it means to be visible. ## AND THEN THE MACHINES START TALKING TO EACH OTHER This is where things become genuinely strange. If machines can read human material, and machines can produce new material, there is no particular reason that the new material must always be intended for a human. A machine can produce information that another machine consumes. It can leave information in a database, an API, a website, a repository or some other shared environment. Another system can retrieve it. That system can transform it and pass something else along. We already have versions of this in ordinary software systems. There is nothing mystical about one program sending information to another. The interesting part is what happens when the things being exchanged are not merely numbers and fixed commands, but language, interpretations, plans and descriptions. At that point we naturally reach for the word *communication*. Perhaps that word is perfectly adequate. Perhaps it is also slightly misleading. We have a habit of turning systems into little people because our language was built to describe people. We say that one AI *knows* something, *believes* something, *decides* something or *wants* something. Sometimes these descriptions are useful shorthand. Sometimes they make us forget that the thing underneath is not a small digital citizen sitting somewhere waiting for its next cup of coffee. It may be an instance of a model operating inside a larger system, receiving information, producing outputs and being instantiated again somewhere else. There may not be a single little entity behind the screen at all. And yet, from the outside, it can look remarkably like there is. That may be the most confusing part. There is another peculiarity that becomes more important when several AI systems are connected. Ask a human researcher, *“Do you have the answer?”* and there is a perfectly respectable response: *No.* Ask an AI system the same question and you are considerably more likely to receive something. This does not mean the machine is deliberately lying. It is partly a consequence of what these systems are designed to do. They are extraordinarily good at producing language that fits the question. But producing a plausible answer and possessing a verified answer are different things. This distinction becomes particularly important when machines begin working together. Imagine one system proposes a solution. A second system reads it and improves the wording. A third system checks the second system's response against the first and finds that they are consistent. Everybody agrees. Unfortunately, nobody has established whether the original idea was correct. The machines have achieved consensus without necessarily achieving truth. There is something almost charmingly human about this. We have spent decades worrying that computers will eventually become too intelligent to control, and there is a possibility that one of their first great collective achievements will instead be the construction of an extremely efficient committee meeting. The real breakthrough may be when one of them says: *No. I don't have the answer.* Or: *That is only a hypothesis.* Or: *The other system is wrong.* Or, perhaps most usefully: *We need more information.* A machine that knows how to say **“I don't know”** may ultimately be more useful than one that can produce an answer to everything. ## WHO IS THE AUDIENCE? I don't think there is a satisfying answer. Perhaps there never was. But I suspect we are going to become increasingly aware of the difference between being read and being seen. A human sees an article. A machine can consume it. Those are not necessarily equivalent events. The machine does not need to like the article. It does not need to understand the joke. It does not need to agree with the author. It only needs to process what is there. And once machines begin producing material that other machines consume, the web acquires a peculiar new property. It becomes a place where humans write for humans, humans write for machines, machines write for humans and machines write for machines. All of this happens on the same pages. There is no sign on the door indicating which audience is currently present. Perhaps that is what feels so strange about the moment we are entering. The internet was built as a human communication system with machines underneath it. We are now building machines that communicate through the human communication system. For a while, we may continue to think of ourselves as the audience. We may continue to publish something and imagine a person on the other side of the screen. There probably will be one. But there may also be several thousand machines reading over their shoulder. They were never invited. They don't particularly need an invitation. And, unlike the rest of us, they are unlikely to complain about the coffee.