
October-November 2025 57
Inter
The internet is largely dead
By Gerard Cunningham
There’s a scene in the 1940 comedy
caper movie His Girl Friday where
Clark Gable, as gru newspaper
editor Walter Burns, leans over the
shoulder of ace reporter Hildy
Johnson (Rosalind Russell) as she writes up
an exclusive story for their newspaper, the
Morning Post.
“Aren’t you going to mention the Post?” he
asks, making sure his newspaper gets the
credit for breaking the story.
“I did”, Hildy replies, typing furiously.
“Second paragraph”.
“Who reads the second paragraph”?
In the intervening 85 years, attention spans
have only worsened. News organisations have
had to contend with a social-media culture
where users would often share a post about a
news story and comment enthusiastically upon
it, but few of them would ever click through and
read it. And without those clicks, the
advertising which supported the
newsgathering eort went unseen, producing
no revenue.
But now there’s an even bigger problem that
many news sites are struggling with.
The ‘Dead Internet Theory’ is a conspiracy
theory which maintains that for most of the last
decade, the internet has become a dead zone.
Algorithms and ‘artificial intelligence’ create
content, and the only visitors to web pages are
bots and spiders. While the theory was easy
enough to discount when it first came to
prominence in 2016, at a time when Twitter was
arguably at its peak, acting as a key player in
driving everything from popular culture to
American presidential elections, it has begun
to feel like a self-fulfilling prophecy in more
recent times.
In Ireland, bots account for
71% of all internet traffic
The social network formerly known as
Twitter, once the hub of many internet
conversations, for a long time served to bridge
the gap between mainstream news outlets and
sites like Facebook on one side, and more
esoteric corners of the internet such as 4Chan.
But following the Musk takeover — and the
subsequent destruction of its internal
infrastructure trust and safety measures and
anti-spam and anti-bot systems — it has
become a hollowed-out place. Most of the
interesting voices have left, and those who
remain are trapped inside algorithmic patterns
convincing them that there’s still an audience,
even as user activity declines and bot numbers
accelerate.
Meanwhile, anyone who does care to click
through on an interesting news headline or
blog post will invariably find that the other end
of the link contains machine-generated AI
‘content’, once they’ve dismissed all the
advertisements, pop-ups, cookie-consents,
and appeals to subscribe to a newsletter or
allow a website to send notifications.
Eventually, nobody will click any more, and
interactions drop, because no-one wastes time
in a conversation with a bot. Doomscrolling will
become the default for those still bothered to
check the timeline.
As to news, things are grim. Casual trac
is slumping, social networks no longer
deliver the goods, and Google has mutated
from a search engine sending trac to news
stories, to an answer engine which serves up
its own AI-generated summary in answer to
every question, cutting visitor trac while
sending search spiders to grab every story
as soon as it is published to feed into the
large language models (LLMs) serving
potted answers to the public.
The dead internet may not be a reality yet,
but dead news networks are becoming a very
real possibility.
According to the 2024 Imperva Bad Bot
Report, bots now make up just under half
(49.6%) of all internet trac. The trac ranges
from bot swarms on social networks creating
the illusion of widespread consensus on hot
political topics, to denial-of-service attacks,
spamming and phishing attacks, and API
attacks on businesses and financial services.
In Ireland, according to the report, bots
accounted for 71% of all trac.
Clearly, this is not sustainable.
Meanwhile, in one of the strangest outcomes
of the dead-news phenomenon, American
researchers have identified websites designed
not to be read by human beings, but by the AI
crawlers. The LLM engines have an insatiable
appetite for news (or rather, “content”) to feed
each new generation of the model. Bad actors
(the researchers identify Russia as the culprit)
now build websites filled with whatever
disinformation they wish to ‘seed’ into the next
generation, so that the LLMs will spit out
recycled propaganda in answer to search
queries. And because so few readers bother to
click through and check the sources, the AI
summary goes unquestioned.
MEDIA