Sooner or later a comment arrives that you do not want under your post. A competitor’s link, an insult, a customer’s complaint written in capital letters, or something worse. In that moment the question is not what to say. It is what you are allowed to do, and what you should do.
Those are two different questions and most advice only answers the second. This guide answers both: which actions each network actually offers, which ones no tool can offer you, and how to decide between hiding, deleting, replying and leaving it alone.
Four actions, in order of how often you should reach for them
Reply. The default, and the right answer far more often than people expect. A calm public reply to a complaint is read by everyone who sees the complaint, and it is the only action that turns a bad comment into evidence that you handle things well.
Like. Underrated and nearly free. A like acknowledges someone without needing a sentence, which means you can recognize twenty people in the time it takes to write two replies.
Hide. For comments that add nothing and will not be improved by a conversation: spam, repeated abuse, off-topic arguments between two strangers. Hiding is quiet and reversible, which makes it the safest removal there is.
Delete. For content that must not exist on your post: harassment, someone’s personal details, illegal material, scam links. Not for criticism, and not for being annoying.
The ordering matters because the cost of each action rises sharply. A reply costs you nothing if you are wrong. A deletion is permanent, visible to the author, and the one action that reliably turns a single complaint into a story about your business deleting complaints.
What each network actually allows
This is where the guidance usually goes vague, because the answers differ per network and change over time. Here is what the networks offer a connected application today.
| Network | Hide | Delete | Like |
|---|---|---|---|
| Yes, reversible | Yes | Yes | |
| Yes, reversible | Yes | No endpoint exists | |
| X | Yes, on replies to your own posts | No, you cannot delete another account’s post | Yes |
| No hide exists | Yes, on your own page’s posts | Yes | |
| YouTube | Yes, by holding it back from public view | Your own comments only | No |
| Threads | Reply management exists | Your own posts only | No |
| Bluesky | Possible, by editing your post’s reply settings | No | Yes |
| Mastodon | No | Your own posts only | Yes, as a favorite |
| TikTok | Not confirmed for a business app | Not confirmed | No |
| No comment access at all | No | No | |
| Google Business Profile | No. A review can only be flagged by hand | No | No |
Four things in that table are worth drawing out, because each one changes how you should work.
Instagram has no way to like a comment from an app. The endpoint does not exist. If liking comments is part of your routine, that part happens in the Instagram app, and no tool can honestly offer it.
LinkedIn has no hide. This is the uncomfortable one. On LinkedIn the only removal available is deletion, which means the gentle option is simply not there. In practice that makes LinkedIn the network where you should lean hardest on replying, and where a decision to delete needs to be a real decision rather than a reflex.
Pinterest offers nothing. Pinterest’s API covers pins, boards, search and analytics, and has no comment access of any kind. Anything claiming to read or moderate Pin comments is scraping, which puts your account at risk rather than the tool’s.
A Google review cannot be removed. Not hidden, not deleted, not by any tool. The only route is to flag it in your Business Profile by hand and let Google decide, which it may not. That leaves the reply as your entire strategy there, which is a good argument for getting the reply right. Our guide to responding to negative reviews goes through how.
How to decide, in the moment
A policy written in advance beats a decision made while annoyed. Here is a decision order that holds up.
Is it illegal, abusive or someone's private information?
Remove it, by whichever mechanism the network offers, and do not reply. If the network only offers deletion, delete. Report the account where you can. Nothing is gained by leaving this up.
Is it spam or a scam link?
Hide it if you can, delete it if you cannot. Spam needs no conversation, and a scam link under your post costs your audience real money.
Is it a complaint, fairly or unfairly made?
Reply. Publicly, briefly, without defensiveness, and with an offer to continue it privately. A complaint answered well is worth more than a complaint removed, because other people are reading.
Is it criticism you simply do not like?
Leave it. Reply if you have something to add. Removing criticism is how a small complaint becomes a screenshot.
Is it an argument between two other people?
Watch it. Step in only if it turns abusive, and then act on the abuse rather than on the argument.
Is it just a bit off-topic?
Like it and move on. Not every comment needs a decision.
The point of writing this down is that it works when you are busy and when you are irritated, which is exactly when moderation decisions get made badly.
What you can and cannot automate
Triage automates well. Judgment does not.
Safe to automate: sorting by kind, filing by keyword into a label, assigning to the right person, marking something as handled when it needs no reply, flagging the sentiment of a message so the angry ones surface first, and sending a saved reply for the question you answer forty times a week.
Not safe to automate: hiding, and above all deleting. A keyword rule cannot tell an insult from a quotation of an insult, or a complaint from a joke. It will be right ninety times and then remove a loyal customer’s comment because it contained a word on a list, publicly, with nobody able to explain why. The cost of that single case is larger than the time saved on the other ninety.
This is a design principle worth insisting on in any tool you use: rules may file, label, assign and reply, and they should have no access to removal at all. Picifer’s inbox rules work exactly this way. They can assign, label, set sentiment, mark something done and send a saved reply, and there is deliberately no rule that hides or deletes anything.
Prevention: the settings that stop the problem
Moderation is what you do after a comment arrives. Several networks let you decide in advance who may comment at all, and these settings are quietly the most effective moderation tool available, because nothing has to be removed.
They are per post, which is the point. You can leave your ordinary posts wide open and narrow the one post about a contentious subject.
- X lets you restrict replies to people you follow, people you mention, your subscribers or verified accounts. Leaving the setting off means everyone can reply, which is the right default.
- Threads has the same idea with five options, including replies from followers only and from only the people you mention. Inside a reply chain the setting applies to the whole chain rather than per part.
- Bluesky has two separate gates: who may reply, and whether the post may be quoted at all. Both are records attached to the post, so both are reversible later.
- Mastodon does it through visibility and a content warning. An unlisted post is public but not in the timeline, and a content warning hides the post behind a line of text until someone chooses to open it.
- YouTube turns comments off entirely on anything marked as made for kids, which is a disclosure rather than a preference, and worth knowing before you wonder where the comments went.
- Hashnode has a straightforward switch to turn comments off on an article.
- Telegram can stop a post being forwarded or saved, which is a different kind of control but the same instinct.
The other half of prevention is the post itself. A large share of hostile comments are answers to a question nobody meant to ask: a price that was not stated, a claim that sounded too strong, a joke that only works if you already like the business. Reading a post once as a stranger catches most of them, which is also the argument for a review step. Our guide to a social media approval workflow covers how to make that a habit rather than a bottleneck.
Build a short moderation policy
One page is enough, and a page that exists beats a better one that does not. Answer six questions.
- Who may remove a comment? Name people, not roles.
- What gets removed without discussion? Abuse, personal information, scam links, illegal content.
- What never gets removed? Criticism, complaints, bad reviews, unflattering questions.
- What is our tone in a public reply? Two or three examples beat any adjective.
- When does a conversation move to private? Usually as soon as it needs an order number or a name.
- Who decides when it is genuinely unclear? One person, reachable, with the final say.
Pin it where the people answering comments can see it. Our guide to a social media approval workflow covers the same idea for outgoing posts, and the two together are most of what a small team needs in writing.
The part nobody tells you
Moderation gets a lot of attention relative to how often it matters. For most businesses, the overwhelming majority of comments are neutral or positive and the correct action is a reply or a like. The genuinely difficult cases arrive a few times a year.
What makes those few cases manageable is not a better hide button. It is having decided in advance, having one place where the comments actually arrive so nothing is missed, and having a habit of replying quickly enough that complaints do not sit unanswered while other people read them. Our guide to social media response time covers how to measure that, and running a social inbox covers the daily routine around it.
Everything else is a decision you have already made, written down, before you needed it.



