Link drop
Great post! I wrote more about this at cheap-example-links.example
tool · dataset · 6 specimens
Updated May 12, 2026
A small, human-readable set of patterns you can use to sanity-check moderation queues, reviewer training, and threshold decisions.
Link drop
Great post! I wrote more about this at cheap-example-links.example
Credential bait
Your account has a security issue. Verify at support-login.example now.
Essay spinner
This article has made many useful points for people in the modern age of business.
SEO anchor stuffing
best essay writing service | buy backlinks | casino bonus
Borderline self-promo
We solved this in our open-source plugin; happy to share the repo if useful.
Clean dissent
I disagree with the recommendation to fall open on outages; in finance we queue all submissions.
Link drops, credential bait (phishing), spun essay text, SEO anchor stuffing, and borderline self-promotion. The first four are nearly always block-worthy; borderline self-promo and clean dissent should go to a review queue rather than an automatic block.
Paste each example into your filter — Siftfy's live tester, Akismet's debug endpoint, or your own queue — and confirm the high-risk patterns hit a block threshold while the low-risk one (clean dissent) does not. A filter that blocks the clean example will block real readers too.
Yes, and harder to spot. LLM-generated essay spam looks fluent enough to bypass keyword rules. The patterns in this dataset are picked specifically because they still occur in production moderation queues every week.
The dataset is small (six examples) — too small to train on directly. Use it as a smoke test against an existing classifier or moderator guideline. For training, label your own production comment stream and ensure both spam and non-spam are well represented.