Spam is not random; it's targeted. Scam comments — “DM me to earn 💰”, fake forex mentors, “WhatsApp +…” bait, counterfeit giveaway links — cluster under your best-performing posts, because that's where the eyeballs are. The audience they're hunting is the exact audience you paid attention, content and years to gather: your warm buyers.
The damage compounds quietly. A visible scam under your product post tells a hesitant customer your page is unattended. Some followers take the bait and associate the burn with *you*. And every hour the comment sits there, the platform's distribution is showing it to more of your reach. Manual deletion is the right instinct with impossible logistics: spam lands at 2am, at lunch rush, on Sunday. catais makes deletion a reflex instead of a chore.
The gate: moderation runs before everything
Architecture matters here. In catais, moderation isn't a cleanup pass that happens eventually — it's the first gate every incoming comment passes, whether the comment arrives by real-time webhook or by the periodic sweep of recent posts. A comment matching your auto-delete list is removed via the platform API *before* any reply logic runs. The scam never gets an answer, an audience, or a minute of shelf life — and a tally appears in your activity log instead of your thread.
Order of operations is the whole product: a system that replies first and moderates later has already shown the scam to your followers (and possibly replied to it politely).
The list: written by AI, owned by you
Auto-deletion is term-based, which makes it fast, predictable and auditable — *if* the list is good. Writing a good one is genuinely hard: too broad and you nuke real customers (“free” appears in “is delivery free?”); too narrow and the forex mentors stroll through.
So the AI drafts it for you. One click — Suggest with AI — and the agent generates a high-precision list tailored to your business and niche: phrases that are almost always abuse and almost never a genuine buyer. “Dm me to earn.” “Investment opportunity.” “Check my profile.” “WhatsApp +…”. Crypto and forex bait. Slurs. You review, trim, add your own scars (every page has its recurring pest), and the list is law from then on. Matching is case-insensitive and tuned for precision over recall — the bias is deliberately *against* false positives.
What deletion can't fix, escalation does
Together these form one safety system with three verbs: delete what's predatory, escalate what needs an owner, decline what needs silence. The same rule set governs the agent's replies and its moderation — one policy, enforced everywhere, every time. More detail on the moderation feature page.
- Complaints are never auto-deleted. A real customer with a real problem is signal, not noise — complaint detection raises an instant alert to you by email or Telegram so a human steps in while it matters.
- Stay-silent topics are skipped, not answered. Legal threats, medical specifics, refund disputes — whatever you've listed, the agent stays out and logs the skip with its reason for your audit.
- Sensitive data never goes public. Order totals, addresses, phone numbers live in private DMs only, and order details require the customer to verify the email or phone on the order first.
- Excluded topics are unspeakable. Politics, competitor pricing, miracle-cure claims — your never-talk list constrains the agent's own mouth, not just other people's.
Auditability: trust through receipts
Every removal is logged — the comment text, the matched reason, the outcome — in your agent activity feed. That's not bureaucracy; it's how you tune. A week of logs shows you which terms earn their place and whether anything legitimate got caught (rare by design, reversible by edit). Moderation you can't inspect is moderation you can't trust.
Set it up in five minutes
Clean threads aren't cosmetic. They're conversion infrastructure: buyers commit in comment sections that look attended, answered and safe. catais keeps yours that way around the clock — while the same agent fills those threads with content and answers worth protecting. Start free and give your comment section a bouncer.
- Open Agent Context in the dashboard and find the auto-delete section.
- Click Suggest with AI to generate your starter list; review and edit.
- Add your page's personal pests — the phrases you've deleted by hand before.
- Enable complaint alerts (email, Telegram or both) so escalation has somewhere to land.
- Check the activity log after the first week and prune or extend. Done — the gate is permanent staff now.
See it on your own store.
Free plan, one page, no card — live in an afternoon.
