“Dead internet theory” & why it’s becoming more real by the day

You may have heard the term “dead internet theory” doing the rounds on social media in recent days, usually coming from users who are growing increasingly concerned or tired of the influx of AI-generated content online.

The term is both a conspiracy theory and an observable phenomenon. The “theory” posits that now large swathes of the internet consist mainly of traffic and data generated by bots and generative AI, with organic human traffic and usage slowly being drowned out by these automated software.

An article from The Atlantic traces the theory’s likely origin to image boards on the cusp of the clear net, the likes of 4chan and Wizardchan. The same article points to the more extreme versions of the theory, that we are being inundated with bots due to a “government psyop”. We don’t agree with this either but rather believe it’s due to human users organically relying on AI more and more.

You can see examples of the “dead internet” beneath viral posts on the former Twitter, now X, where automated accounts will post and repost content made to generate interest and engagement. Some of these threads won’t see a human poster for a good while thanks in part to X’s current algorithm favouring premium accounts and engagement farms.

Posts like these go on and on and on.

Sometimes the replies from accounts that aren’t the original poster will be on topics that are unrelated, and some users are starting to take notice. Other examples on X include seemingly normal accounts ousting themselves as bots due to errors. Sometimes these can emerge in their hundreds as seen in the Tweet below, descrying dead internet theory on the social media platform.

A dying internet, filled with AI and bots, slowly but surely reduces the novelties and usefulness of the net itself, the same usefulness that made it the beacon of everyday human life that it has become now. Apart from anecdotal evidence, some recent studies are also piling credence to the theory.

Dead internet theory – some recent evidence

Vice covered a study conducted by Amazon Web Services AI lab, which found that a “shocking amount” of the internet is machine-translated gibberish, particularly of content in languages spoken in Africa and the global south ie. languages that exist in the minority online.

“We actually got interested in this topic because several colleagues who work in [machine translation (MT)] and are native speakers of low resource languages noted that much of the internet in their native language appeared to be MT generated,” explained Mehak Dhaliwal, a former applied science intern at AWS and current PhD student at the University of California, Santa Barbara, in the Vice piece.

“With that said, everyone should be cognizant that content they view on the web may have been generated by a machine,” the researcher added.

A serious issue with this is that, like all AI-generated content, machine translation is skewed by the human bias of the coders of that machine translation. This means that translated content will be biased.

“Sentences are more likely to have translations in French than a low resource language, simply by virtue of there being much more data in French than a low resource language,” Dhaliwal added. This means that organic content in languages such as isiXhosa (a low-resource language) will receive a worse translation than a high-resource language, like Spanish.

The eventuality of this is that the translated messages lose the original meaning and intent. This leaves vast amounts of what is essentially rubbish text that doesn’t mean anything floating around online in the megabytes and gigabytes, and this issue will surely only grow. Users that speak low-resource language then have to deal with a much worse internet that users that speak high resource languages, creating a gap in the usefulness of the internet – a type of online classism.

A 2024 study from a team of German researchers found – after a year of collecting evidence – that the pursuit of search engine optimisation (SEO) has slowly been making Google Search less and less useful. It showed that spam sites are increasingly and constrictingly common on Search, and that a large amount of product-related queries on Google led to “outright SEO product review spam.”

This has caused what they call “an overall downwards trend in text quality in all three search engines,” including DuckDuckGo, Google and Bing.

“SEO is a constant battle and we see repeated patterns of review spam entering and leaving the results as search engines and SEO engineers take turns adjusting their parameters,” the report reads. The researchers say that despite efforts from Google and other search engines to remove the spam, it only marks a temporary difference as new spammers take the old spots.

The mega-corporations are only adding to the fire, spreading the dying of the net

Last year in June we covered a NewsGuard report that found just how large and common content generated by AI is across Search. What they call “Unreliable Artificial Intelligence-Generated News” wesbites are climbing in number year on year. These sites can produce “1 200 articles a day.” Most of the content served by these them is nonsense, or worse still plagiarised from actually reliable sources.

Users are spinning up these AI-generated blogs by the hundreds in order to farm ad revenue from companies like Google. Across the thousands of nothing articles they spin up, it is Google who supplies most of the visible advertising.

We saw more examples of dead internet theory in April, when fake South African Social Security Agency (SASSA) websites were popping up in their dozens, serving pre-generated instructions created on ChatGPT on how people can receive their social grants. Many of these sites ask for private information even when they are not affiliated with official government channels. Nothing can seemingly be done about them.

Every day it seems like more and more material we see online is machine-made, with low-effort content the first to be replaced by AI-generated images, and AI-generated text. What was once a funnel for all of mankind’s insight, history, learning, creativity, and helpfulness is evidently slowly transforming into a mulch – a cistern of nothing content, served fresh by billions of bot networks and eaten up by unsuspecting users, paid for by the megacorps.

Now, real, authentic human thought and word are being dragged to the outskirts of the net as profit-seeking algorithms create the perfect environment for the internet to simply die.


About Author


Related News