Comment Moderation Best Practices

Good moderation removes what hurts the ad (spam, scams, abuse) and keeps what helps it: questions, praise and complaints that get a public answer. Every viewer reads the thread before deciding whether to click.

Hide, answer or escalate

Comment	What to do	How in Rypl
Spam, bot text, link spam	Hide	Automatic; add Strict Spam & Media Filter for ads that draw link spam
Scam links, fake “support” accounts	Hide	Automatic, plus banned words or a rule
Hate, slurs, threats	Hide (or delete)	Automatic, plus banned words
Competitor names and links	Hide	Banned words, or a rule: comments recommending another store → Hide
Questions about price, stock, shipping	Answer	The agent answers by default
Complaints about an order	A person answers	Needs attention by default; a rule can Flag for me
Refund and return requests	A person answers, or the agent with your policy	Agent answers from the knowledge base; a rule can send them to you
Praise	Thank	Thank-you replies (Pro and Ultra) or Like positive comments (Facebook)
A lone emoji, friends tagging friends	Leave	Nothing to set

Hide, don’t delete, by default

A hidden comment stays visible to its author and their friends, so they don’t know and don’t come back angrier. Hiding can be undone; deleting can’t. Keep Delete for comments you never want anyone to see, such as slurs or doxxing. On TikTok, delete isn’t possible for other people’s comments, so Rypl hides them.

Leave legitimate complaints visible

A complaint answered in public shows every later viewer that you respond. A thread with no criticism at all looks curated. That’s why Filter by tone is off by default and complaints go to Needs attention. Turn on Hide negative only for ads that attract hostility unrelated to your product, such as political arguments.

Banned words that don’t catch customers

Use quotes for short words: "ass" matches only the whole word.
Without quotes, a word matches words that start with it: scam also catches scammer.
Use asterisks only when you mean “anywhere”: *casino*.
Add competitor names in every spelling buyers use.
Check Hidden by banned words in the inbox weekly for innocent comments.

Rules you can trust (Pro and Ultra)

Write rules like instructions to a new hire. Hide comments that recommend another store or include a discount code that isn’t ours.
One idea per rule. Two short rules are easier to check than one long one.
Start in Review first. Matches wait for you labelled with what the rule would do. Switch to Apply automatically when the rule is right almost every time.
Write global rules once for what applies to every brand, and add rules Only on this channel for one brand.
Use I’ll answer myself for comments the AI should never touch, such as legal threats or press enquiries.

Review what was hidden

Every week, open the inbox filter All hidden or deleted and skim it. Unhide mistakes, then fix the cause: a banned word that’s too broad, a rule that’s too vague, or a tone filter that’s too strict.