An increasing amount of content and traffic on the internet is created by machines – people are the online minority. The so-called "dead internet" theory is no longer just a meme: generative AI has flooded the web with text, images, and audio. What does this mean for security, trust, and everyday navigation in the digital world – and how do we defend against it?
When the web fills with synthetic content
A substantial share of web traffic today comes from bots and generative models. The explosion of tools for generating text and visuals has made "synthetic" content cheap, fast, and ubiquitous. The dead internet theory thus moves into practice: in the online noise, humans are often drowned out by machines.
The symptoms are visible on social networks: monotonous "SEO" formulas, "rage bait" and "AI slop" boost reach regardless of quality. Automated podcasts can spew out thousands of episodes per week, and estimates say that a noticeable share of new music releases are created with AI. Even old familiar spam tactics are now powered by artificial intelligence, which scales fraudulent campaigns and targets more precisely. The result is a flood of content that is cheap to produce but expensive to verify.
OSINT and identity cloning: a low bar for attackers
Open sources (OSINT) make it possible to piece together a surprisingly accurate profile of a person from public data. Automated search, web scraping, API services, and "the internet never forgets"–type archives turn past traces into a permanent source of information. The barrier to entry has never been lower – just a few tools and a bit of skill are enough.
Face search can find dozens of other images across the web from a single profile photo within seconds. Terms of use won't stop these risks, and OSINT feeds into more sophisticated phishing: precisely targeted messages and fake pages indistinguishable from the original. The next step is voice – one minute of public recording often suffices for a usable clone, which commercial services can further fine-tune with parameters. And on video calls, attackers can deploy deepfakes that blur the line between reality and simulation.
How to defend yourself: verify, slow down, humanize
For suspicious requests, switch channels and escalate verification: confirm a phone call with video, confirm video with a call back to a known number or a message via another channel. Watch for warning signs and artifacts – some deepfakes still fail on unexpected gestures or when part of the face is covered. Slow down decision-making for financial and access requests, and require multi-step confirmation via at least two independent channels. Ongoing education about fraud trends is now as important as antivirus.
In the long term, supporting smaller, human-curated communities and channels with verified authors will help. A fully "humans-only" network is technically and organizationally hard, as distinguishing a person from a bot will become increasingly difficult. It's a cat-and-mouse game: attackers and defenders constantly adapt. Even so, a backlash is likely – fatigue with the glut of synthetics may increase the value of trust, quality, and transparent provenance of content.
