View source for Blog:The Internet Is Dead. Long Live the Internet!

{{Blog header
|Title=The Internet Is Dead. Long Live the Internet!
|Categories=LWP meta, Technology and infrastructure
|Author=Michele Lavazza
|Date=2026-04-26
|Intro=Over the last months, the LWP’s visitors have been mostly bots. On the one hand, this makes our web traffic statistics useless. On the other hand, it raises the question: what sense does it make to work on a free culture web project if the Internet is dead?
|Image=File:Henri II et Catherine de Médicis.jpg
|Status=published
}}
About a year ago, we published [[Blog:LWP readership statistics 2024|a blog post]] about the traffic statistics for the Ludwig Wittgenstein Project’s site. The data we had gathered through our web analytics tool, Matomo, provided a rough measure of the general usefulness of the Project. When looked at in detail, it also showed which languages and individual texts had attracted the most visitors. Alongside other metrics and considerations, it proved useful in shaping our strategy going forward. We thought the data would be interesting to our audience as well, and we decided to make a habit of publishing a yearly readership report.

Nothing would have made me happier than to follow through on that resolution. This post describes the circumstance that made it impossible and draws some reflections from it.

In the 7-month period between January and July 2025, the site had experienced the same steady increase in traffic that it had experienced in the previous 7 months (June–December 2024).

Visits had gone up 40.7% (from 36,158 to 50,859) and pageviews had gone up 31.8% (from 64,459 to 84,931). The bounce rate (percentage of visitors who had left the website after only viewing one page) was stable at 72-73%. The average visit duration had gone up 9.8% (from 1 min 42s to 1 min 52s). Everything indicated a sustained and sustainable growth. 

[[File:Matomo visits over time comparison (late 2024 - early 2025).png|class=img-comfy|alt=Line chart showing the number of visits between late 2024 and early 2025|center|thumb|600x600px|Weekly visits. The orange curve is the 7-month period between June and December 2024; the blue curve is the 7-month period between January 2025 and July 2025.]] 

Then, in the last week of August 2025, we experienced a surge in traffic: from a respectable weekly average of about 1,500 visits around July and early August to the frankly absurd figure of about 50,000 (and up to 93,000) average weekly visits between the end of August and September. All other data (page views, unique page views, searches…) is also inflated beyond recognition after August 2025.  

[[File:Matomo visits over time 2025-01-01 - 2026-01-31.png|class=img-comfy|alt=Line chart showing the number of visits between 1 January 2025 and 31 January 2026|center|thumb|600x600px|Weekly visits. The spike at the end of August 2025 is clearly visible.]]

We were able to make sense of this unexpected change of familiar patterns by looking at the segmentation of the data by geographic location and cross-referencing it with the segmentation by individual pages viewed. We found out that, since September 2025, 95% of the visits had come from IP addresses in China, and that the vast majority of them had not been visits to content pages but to service pages such as [[Special:Templates|Templates]], [[Special:Log|Main public logs]], [https://www.wittgensteinproject.org/w/index.php/Special:RecentChangesLinked?hidebots=1&target=&limit=50&days=7&enhanced=1&urlversion=2 Related changes], [https://www.wittgensteinproject.org/w/index.php/Special:RecentChanges?hidebots=1&limit=50&days=7&enhanced=1&urlversion=2 Recent changes] and so on. The purpose of these visits is unclear: they may be aimed at retrieving unindexed data, compiling some statistic of their own, or looking for security vulnerabilities, which, in turn, would be exploited to try to gain unauthorised access to the website’s backend. Their number and pattern make it quite clear, however, that they are automated: these are not people, but bots.

Luckily, although additional traffic puts additional load on the server, there has not been a noticeable effect on the performance of the website, with the average server time remaining stable at approximately 0.55 s. Human users, in other words, are not experiencing slower load times due to the activity of automated users.

The only practical effect of the traffic soaring, then, is that the site usage statistics no longer make any sense. We are unable to filter the automated accounts that parse our web pages and separate them from the humans who actually read them, and because the vast majority of the visits, page views, etc. are now coming from the former, the numbers are suddenly meaningless. We had 1.2 million visits between September 2025 and January 2026. We estimate that more than 90% of them came from bots, but we do not know which ones, and this makes it impossible to extract reliable reports on the breakdown of (human) visits by language, by country, by text, etc.

On the one hand, this is it. One particular instance of the Dead Internet has rendered the Ludwig Wittgenstein Project’s web analytics tool useless. Frustrating as it might be for us, it carries little import. On the other hand, there is a follow-up thought I think is worth articulating, because it clarifies ''why is it'' that this frustration is not important.

The [[wikipedia:Dead_Internet_theory|Dead Internet theory]] started as a conspiracy theory claiming that most of what is available or happens on the Internet (content, social media interactions, emails, web traffic, etc.) is the result of the activity of bots. In its original form, the theory also argued that behind these bots is a coordinated, malicious intent to control human users. Whereas this latter part of the theory can hardly be taken seriously (especially because of its generalising stance, whereby all or most of automated activity on the Internet is part of a ''single'' deliberate effort to manipulate us), the former part has become increasingly, factually true.

The Dead Internet theory has been around since at least 2021, that is, since before the boom of generative AI, which started with the launch of ChatGPT in late 2022. Now content created by Large Language Models is very much on the rise and, according to [https://graphite.io/five-percent/more-articles-are-now-created-by-ai-than-humans a widely cited study], it is already the case that about half of the new articles published online are AI-generated. On a more subjective note, we have all noticed, albeit surely with varied degrees of annoyance, the proliferation of AI-generated text and images on social media.

The case of the LWP’s traffic is minor, but it is factual and it fits the trend that the Dead Internet theory captures.

Overall, the Internet is now a rather bleak place. Partly because of the AI slop flood, and partly because of the phenomenon for which Cory Doctorow coined the name “[[wikipedia:Enshittification|enshittification]]”. Big platforms are past the phase in which they were trying to lock in users by offering a good experience. They are now pursuing strategies directing every effort to the monetisation of the time and attention of users that have already been locked in, and the users’ experience is very much sacrificed in service of whatever creates the highest degree of engagement and, ultimately, addiction.

If one thought that the Internet must deserve saving in order for one’s efforts to improve the Internet to be meaningful, then one might easily give up those efforts: the deader, the more enshittified the Internet is, the less it deserves saving. Yet, precisely when we look at it this way, it becomes clear, I think, that what we do to cultivate some small corner of the Internet, and make it look a little like a public park, a little like a public library—what we do to this end is meaningful not because the Internet deserves it, but because it needs it. Because ''it needs it'' to be, at least in part, what it was meant to be in the early, idealistic days, and what we still want it to be if it is to serve humankind as best as possible. Because, in other words, ''we need it''. 

From the point of view of the LWP it is easy to feel discouraged because on a superficial level, by looking at the mere numbers, one has the feeling that one’s audience is 90% made up, which in turn feels almost the same as talking to an empty room. But then one realises that the 10% of users who are not made up are the only ones that matter. It is then that the work looks worthwhile again.

The TV did not kill the radio, the Internet did not kill the TV. In a similar way, the enshittified Internet did not kill the free Internet (the cost-free, ad-free, surveillance-free and libre Internet). The dead Internet did not replace nor displace the Internet that is alive. Although there is indeed more AI slop and more bot noise in general, the places where there is neither—where humans find valuable content curated by other humans—are still there and are there to stay. Wikipedia, the Internet Archive, Project Gutenberg and the others, including our very own tiny LWP, are playing the long game. The Internet is dead: long live the Internet!

{{p right|<small>Cover image:  [https://commons.wikimedia.org/w/index.php?curid{{=}}11209266 Henri II et Catherine de Médicis] by Germain Pilon, CC BY-SA 3.0.</small>}}
{{Blog footer}}