AI slop is killing online communities

Original Article ↗ Hacker News Discussion ↗

Perceived Rise of “AI Slop” and Bots

  • Many see a sharp increase in low-effort, LLM-generated posts and comments across Reddit, HN, YouTube, etc., often indistinguishable from mediocre human content.
  • Experiments with karma‑farming and covert advertising on Reddit convinced some that readers can’t tell they’re talking to bots.
  • Others think most obvious “slop” is still easy to spot and that the flood is driven as much by bots, marketers, and propagandists as by casual users.

Quality and Detectability of LLM Content

  • One side: LLM writing is shallow, formulaic, logically weak; good at surface polish but collapses under scrutiny.
  • Other side: with careful prompting and style constraints, current models can reliably “pass the sniff test” in casual online discourse.
  • Some complain that people now cry “AI slop” at any comment they dislike, turning it into a new form of lazy dismissal or witch‑hunt.

Impact on Platforms and Communities

  • Reddit is widely described as “cooked”: bot reposts, fake reviews, karma farms, political and commercial astroturfing, weakened moderation tooling, and leadership incentives that favor fake engagement.
  • HN is seen as under pressure but better moderated (Show HN filtering, autokilling some LLM comments), though some worry about front‑page AI-written blog posts and subtle influence campaigns.
  • Discord, Facebook groups, mailing lists, and small forums are reported as healthier, especially when private, invite‑only, or tightly moderated; public Discords are already seeing similar problems.

Moderation, Gating, and Reputation

  • Proposed defenses: paywalls or small signup fees, strict invite/vouch trees, web‑of‑trust models, slow modes for newcomers, heavy banning of AI content, and domain/community‑level curation.
  • Critics note: money and invites can be gamed or bought; reputation systems and votes are easily manipulated; strict gates reduce accessibility and diversity.
  • Some moderators already outright ban AI‑generated content and say early, firm lines work better than “wait and see.”

Identity, Proof‑of‑Human, and Privacy

  • Many argue public communities will die without some form of human attestation: government ID, verifiable credentials, ZKPs, device/TPM‑based proofs, or Worldcoin‑style schemes.
  • Counterarguments: IDs can be stolen, rented, or used by humans who then delegate to bots; centralized ID is a huge privacy and abuse risk; compromised devices could mass‑create “real” spam accounts.
  • Web‑of‑trust ideas (PGP‑style, Debian‑style key signing, invite trees) are seen as promising but inherently exclusionary and vulnerable to infiltration.

Online vs IRL and Small‑Scale Communities

  • Many predict a shift from large anonymous publics to:
    • Smaller, invite‑only or paid communities.
    • Mixed online/IRL networks (clubs, meetups, games, voice/video).
  • Some hope AI slop will push people back into physical “third places”; others note online niches remain essential for rare interests and geographically isolated people.

“Does It Matter If It’s AI?”

  • One camp: if you can’t tell and the content is useful, source doesn’t matter.
  • Opposing view: it matters because people seek human perspectives, want to help real humans, and rely on human judgment; talking to bots erodes trust and makes participation feel pointless.

Long‑Term Outlook

  • Several see this as a “tragedy of the commons” moment: without new mechanisms (economic, social, or technical), open public forums will become noisy, manipulable slop fields.
  • Others are cautiously optimistic that:
    • Over time, norms and tools (better search/discovery, slop filters, reputation systems) will adapt.
    • High‑effort, high‑signal communities will survive behind stronger gates, even if the wide‑open web decays.