Ask HN: Add flag for AI-generated articles

Hacker News Discussion ↗

HN users are debating whether links to AI-generated or AI-assisted articles should be labeled or discouraged, given that the site already bans AI-written comments. Many argue that AI prose signals low effort, erodes trust, and undermines the “human conversation” ethos, while others see AI as a legitimate assistive tool, especially for people with weaker writing skills, and insist quality should matter more than the tool used. Proposals range from new flag reasons and [AI] tags to browser-side filters and stronger norms against “AI slop,” but concerns about false accusations, stigma, and enforceability remain unresolved.

HN’s Current Rules and Moderator Input

  • Generated or AI-edited comments are banned; software plus human moderation enforces this with mixed success.
  • There is no explicit rule yet for AI-written articles, but moderators see strong community skepticism and “discounting” of LLM prose.
  • Moderation is experimenting with AI-detection; some legitimate comments are auto-killed, and users can vouch or email to restore them.
  • A likely change: require a reason when flagging posts (e.g., spam, off-topic, “genai”), not just a generic flag.

Proposed AI Flags/Tags and Filtering

  • Suggestions include:
    • A specific “AI-generated” flag reason.
    • Visible tags like [AI] / [LLM] or even [SLOP].
    • User settings to hide posts with many AI flags.
    • Community tagging, or “not AI” / “handmade” labels.
  • Others propose browser-side solutions (uBlock blocklists, extensions that filter AI domains or keywords).

Arguments for Restricting or Labeling AI-Generated Articles

  • Many see AI-written posts as low-effort “slop” that breaks the implicit effort balance: if the author didn’t invest, why should readers?
  • LLM prose is perceived as formulaic, overblown, and often inaccurate or hallucinated, wasting readers’ time.
  • AI authorship undermines discussion: it’s unclear what the human actually knows, so feedback and deeper questions become pointless.
  • Some want HN to remain a human-to-human space and fear broader “dehumanization” and loss of genuine voice.
  • Labels help readers who explicitly want to avoid AI content and protect limited “cognitive bandwidth.”

Arguments Against AI Flags / Stigma and Abuse Risks

  • Detection is unreliable; false accusations of “AI slop” are seen as insulting and corrosive, already causing drama elsewhere.
  • “AI-generated” is becoming a slur used to dismiss work or viewpoints, sometimes without evidence.
  • Users describe ableist dynamics: AI can be assistive tech for people with ADHD, dyslexia, or limited English; blanket stigma marginalizes them.
  • Honest disclosure of AI involvement could be punished if the label lowers status; liars would be rewarded.
  • Some consider AI a tool like spellcheck or word processors; they care about informativeness and rigor, not the toolchain.

Detection Challenges and What Counts as “AI-Generated”

  • Human detection is inconsistent; some are confident, others admit many misclassifications.
  • Dedicated AI detectors are also fallible; posts note both false positives and model limitations.
  • Boundaries are fuzzy:
    • AI as copyeditor vs full author.
    • AI-generated illustrations or sections in otherwise human text.
    • AI-assisted research vs AI-fabricated citations.
  • Several argue that only obvious, unedited “LLM voice” is currently easy to spot; future models and “humanizers” may blur the line further.

HN Voting, Flagging, and Community Dynamics

  • Regular voting is seen as insufficient because submissions can’t be downvoted, only flagged.
  • Some users already flag AI-looking submissions as “not belonging on HN,” despite no explicit rule.
  • Concerns that flags and downvotes are often used to punish unpopular opinions, not just rule violations.
  • Suggestions include richer voting systems (quality vs agreement axes) and clearer guidance on legitimate uses of flagging/vouching.

Broader Reflections on AI, Writing, and the Web

  • One camp: only quality matters; a great article is great regardless of origin.
  • Another camp: authorship and “proof of work” matter for trust, human connection, and status; they explicitly seek non-AI writing.
  • Some predict the majority of web content will become AI-generated; in that world, high-quality human work becomes the rare signal.
  • There is tension between embracing AI as a powerful coding and writing assistant and preserving spaces optimized for authentic human communication.