# Social Media Comment Moderation for Facebook & Instagram | Zipchat Hide spam, scams and abuse under your Facebook and Instagram posts and ads automatically, before shoppers read them. Works on dark posts. ## TL;DR Zipchat's AI reads each comment under your Facebook and Instagram content and decides what to do with it: hide it, answer it, or leave it alone. It runs on organic posts, boosted posts and unpublished dark post ads, without keyword lists to maintain. ## Key facts - Hides spam, scams, competitor links and abuse across Facebook and Instagram - 36% of all agent actions in early access were hiding a comment - 1,558 comments hidden across 33 brands in the first three weeks - Works on organic posts, boosted posts and unpublished dark post ads - Median 13 seconds from comment to action - Hidden comments stay visible to their author, who receives no notification ## Can AI hide spam and abusive comments on Facebook and Instagram? Yes. Zipchat connects to your Facebook Page and Instagram account, reads every incoming comment, and takes an action on it within seconds. Unlike a keyword blocklist, which only catches the words you thought of in advance, the agent reads the comment in context and decides whether it is a real customer, a scam account, a competitor, or abuse. The distinction matters more than it sounds. A blocklist that hides the word "refund" also hides the customer who is asking a fair question about one. Zipchat separates the two, because it understands what the comment is asking for and it has your order data behind it. Why this is a revenue problem, not a hygiene problem. A Harvard Business School working paper ran two field experiments on live Facebook and Instagram ad campaigns. When automated moderation was switched on, return on ad spend rose from 0.459 to 0.678 and cost per purchase fell from $666.56 to $454.34. In a controlled follow-up with 2,485 participants, hiding harmful comments raised stated purchase intent from 36.9 to 47.2, a significant effect. Replying to comments, tested on its own in the same experiment, did not produce a significant effect. That result is worth sitting with. The people affected by your comment section are mostly not the people writing in it. They are the shoppers reading it before they decide. Peer-reviewed work in the Journal of Marketing Research found a single visible negative review on the first page cut purchase probability by 41.8%. A comment section under a paid ad works the same way, and you are paying to put it in front of people. Does moderation cost you reach? No. The same Harvard experiment measured impressions and reach across both arms and found no meaningful difference. Delivery was unaffected. What changed was whether the traffic converted. Zipchat handles the public layer and then continues the conversation privately through comment to DM automation, or hands it to a person with human chat takeover. For the full picture of what the agent does with a comment, see the AI social media manager overview. ## Scenario Description ### Spam Under a Live Ad Campaign A supplement brand launches a Meta campaign on a Friday afternoon at $2,000 a day. Within an hour the ad has ninety comments. Eleven of them are scam accounts posting "message this number for 70% off", using a display name close enough to the brand's own that a shopper skimming the thread would not catch it. Four more are competitors dropping their own link. Nobody on the team is looking. It is Friday evening, the campaign runs through the weekend, and the ad keeps serving. Every impression the brand pays for on Saturday shows that comment section underneath the creative. With Zipchat connected, each of those fifteen comments is hidden within seconds of being posted. The scam accounts see their own comments still sitting there, so they do not know to try again from a different handle. Meanwhile a real comment in the same thread, "is this safe with blood pressure meds?", gets a public reply pointing to the ingredient page and a private message opening a proper conversation. Why moderation pays for itself here: the ad spend is the same either way. The only variable is what the next shopper reads before deciding. ### A Complaint That Should Not Be Public A customer comments under a product reel: "third week waiting, no tracking, this is a joke." It is a real complaint from a real customer, and hiding it would be the wrong call. Deleting it would be worse, because deletion is conspicuous and this person is already angry. Zipchat is set with a rule the brand wrote in one sentence: never discuss a delivery problem in public, always move it to DM. So the agent posts a short public reply, "That is not good enough, we have sent you a message to sort it out", and opens a direct message. In the DM it looks up the order, finds the carrier exception, and either resolves it or escalates to a person with the full thread and the order attached. The public thread now shows a brand that answered in seconds. The problem gets handled where it belongs. And the next shopper scrolling that reel sees a company that responds, not a company with an unanswered complaint sitting at the top of its comments. Why hiding was the wrong tool: moderation is not about silencing criticism. It is about removing what has no business being there, and routing what does. ## Setup Guide - Connect the channel in Channels: Facebook Messenger for your Facebook Page, Instagram for your Instagram account. Comment management runs on the same connection as your DMs, so an active channel is the only prerequisite. - Open the Comments row that appears inside the channel card once the channel is active. - Choose the actions the agent may take. Like, Hide and Ignore are on by default; Delete stays off until you enable it. Each action carries its own instructions, so you decide when it applies. - Write the moderation rules as plain instructions on each action. Example for Hide: "Hide competitor links, discount-code spam and anything abusive." - Set the private reply instructions for what moves to a DM, and the public acknowledgment the agent posts under the comment. The default reads "We've sent you a private message 💬". - Optional: auto-delete hidden comments after 1 to 30 days (the default is 3). Toggle the Comments row on and the agent handles new comments immediately, including under posts and ads already live. The setup is identical for Facebook and Instagram. One platform difference: an Instagram private reply opens a DM thread, and Meta allows one private reply per comment within 7 days. ## FAQ ### Does hiding a comment reduce my post's reach? No. A Harvard Business School field experiment ran the same ad campaigns with and without automated moderation and measured impressions and reach across both. The difference was not meaningful. Delivery is not affected by what you hide. What changed in that experiment was conversion: return on ad spend rose from 0.459 to 0.678. ### Will the person know their comment was hidden? No. On both Facebook and Instagram, a hidden comment stays visible to the person who wrote it and to nobody else, and neither platform sends a notification. This is why hiding is usually the better tool than deleting. A deleted comment vanishes conspicuously, which tends to provoke the kind of commenter you least want to provoke. ### Does this work on comments under paid ads? Yes, on both platforms, including boosted posts and unpublished dark post ads. There is no separate configuration. Ad comments are where most brands need this most, because the brand is paying to distribute the creative that the comment section sits under. ### How does it tell a real complaint from abuse? It reads the comment rather than matching it against a word list. A keyword filter set to catch "refund" will hide a legitimate question about your returns policy. Zipchat separates a customer with a problem from an account attacking you, then applies the rule you set for each. Real complaints get answered, publicly or privately, based on your instructions. ### What happens to comments the AI is unsure about? They go to a person. Anything the agent cannot resolve with confidence is escalated with the full comment thread attached, so whoever picks it up has the context without going to find it. You can also set categories that always escalate regardless of confidence, which most brands do for anything legal or medical. Comment moderation is the part of social media management that nobody enjoys and everybody postpones. It is also the part that a Harvard field experiment measured a return on ad spend improvement from. Both of those things are true at once, which is why it is worth automating first.