The Internet Is Full of AI Dogshit

Original Article ↗ Hacker News Discussion ↗

Widespread use of large language models and SEO automation is rapidly flooding the web with low‑quality, often misleading text, making it harder to find trustworthy human-written information via search engines like Google. Commenters link this trend to the “dead internet theory,” arguing that ad-driven incentives and enshittified search results are pushing people toward closed communities, paid or curated search tools, and ideas like cryptographic signing or reputation systems to verify human authorship. Many see this as an acceleration of long‑standing problems with spam and monetized content, but worry that the sheer volume and plausibility of AI output could trigger a deeper collapse in online trust.

State of the Web and Search

  • Many see the web as having been “full of dogshit” long before LLMs: SEO content mills, affiliate spam, recipe pages padded for ad impressions, clickbait news, and low-quality “how-to” blogs.
  • The sharp change is volume and speed: AI makes it nearly free to flood the web with plausible-looking text and images.
  • Google is a central target: people report results dominated by SEO spam, AI-written explainers that bury the answer, over-aggressive synonym expansion, and ad-heavy result pages that often surface junk over primary sources.
  • Several note that this is about incentives: ad-driven ranking and growth pressure at big firms (e.g., needing tens of billions in new revenue annually) push toward enshittification.

AI Content and “Dead Internet” Concerns

  • Discussion links this to “dead internet theory”: bots and auto-generated content overwhelming organic human activity.
  • LLMs are seen as qualitatively different from old Markov/spintext spam because they produce fluent, on-topic prose that’s hard to spot.
  • Fears include: election misinformation, bot “armies” arguing with each other, AI-to-AI content loops causing “neural network collapse,” and a Kessler-syndrome‑style information junk cloud.

Human vs Machine Authorship and Authenticity

  • Many emphasize that humans have always produced bullshit; the issue is scale and lack of cost, not the existence of nonsense.
  • People report valuing even bad human work (e.g., bizarre personal sites) more than equivalent AI output because there’s a mind and continuity behind it.
  • Proposed defenses: cryptographic signing of human content, identity/attestation systems, webs of trust, and reputation-based filtering. Others note these do not prove “non‑AI” authorship and are vulnerable to key theft and gaming.

Economic Incentives and Enshittification

  • Recurrent theme: advertising and over‑commercialization warped incentives for search, publishing, and social media.
  • AI is seen as a new tool for short‑term profit: mass content generation, SEO gaming, fake reviews, and low-cost “content marketing.”

Proposed Responses and Coping Strategies

  • Technical: new search engines focused on small-web/human sites (Kagi, Marginalia), filters or “whitelists” that downrank likely-AI or ad-heavy pages, client-side tools that summarize/strip junk.
  • Social: retreat to smaller, curated communities (forums, Discords, invite‑only groups), more reliance on books, libraries, and archives (including hoarding pre‑AI web content).
  • Some remain optimistic about AI as a productivity tool (summarizing docs, drafting code) if its output is curated and verified, but others predict deep erosion of trust and expertise.