Spam in conversational replies to blog posts

· ai web books · Source ↗

TLDR

  • A WordPress blog caught a three-comment spam cluster from one Philippine IP, spacing 3 minutes apart, hiding a casino link in the middle reply with no https://.

Key Takeaways

  • The attack uses AI-generated, loosely on-topic comments in a reply chain so the spam sits in the middle, not at the top where review focus lands.
  • No https:// prefix on the embedded link made it invisible as a hyperlink in most comment UIs, bypassing at-a-glance moderation.
  • Antispam Bee blocked hundreds daily but missed this because the social framing (conversation thread, unique emails, no URL fields filled) mimicked legitimate pingback patterns.
  • Same-IP detection and exact 3-minute reply intervals are reliable post-hoc signals but require manual review to catch.
  • Author’s conclusion: adding comment barriers filters real users faster than spammers who have profit incentive to route around them.

Hacker News Comment Review

  • Commenters confirmed this conversational bot-chain pattern is years old on YouTube, with 5-7 fake accounts building a fake discussion thread before dropping a referral link.
  • The WordPress attack surface is specifically targeted: spammers use technology fingerprinting to identify popular blog engines and skip custom stacks with JS-based client-side checks, which see near-zero spam.
  • There is active disagreement on the author’s “no technological solutions to social problems” framing: one counterpoint argues social cohesion alone cannot contain adversarial profit incentives at scale.

Notable Comments

  • @smusamashah: Worked a “link posting” gig in university 20 years ago with Excel sheets of target blogs and started automating it, illustrating how low the barrier to entry has always been.
  • @PaulHoule: XRumer (~2008) already solved CAPTCHAs and email verification for link spam at scale, putting today’s AI-slop chains in a long historical lineage.

Original | Discuss on HN