Title: How to stop spam comments on your Facebook and Instagram ads
This page: https://commentwatchdog.com/blog/stop-spam-comments-on-ads
Topic: Spam
Author: https://anantgupta.xyz (Anant Gupta, founder of CommentWatchDog)
Published: 2026-09-23
Updated: 2026-09-23
Summary: The four filters Meta runs on an ad, in its own words, what each stopped and missed under my ads in September 2026, and the rule set that catches the rest.
Query: how to stop spam comments on ads (20 searches a month in the US)
Tested on: September 5 to 17, 2026, Meta's help pages on the web; September 1 to 23, 2026, the comment log of my own Facebook and Instagram ads; September 8, 2026, my ads in Meta Business Suite
Why this page exists: To answer "how to stop spam comments on ads" for advertisers on Facebook and Instagram with the four layers Meta gives an ad, in Meta's words with dates, what each one stopped and missed under my own ads in September 2026, and the rule set that handles the rest.

In short:
- The short way: connect the Page to CommentWatchDog and its rules hide or delete the spam under every ad on their own, links, phone numbers, emoji strings and pitches included, with the backlog cleared in one run. 7 days free, no card.
- Meta's filters reach ads because they are set higher up. Its words, read September 17, 2026: the profanity filter and blocked words "are set at the Page level and extend to ads from the Page", and Hidden Words does the same for an Instagram account.
- Moderation Assist covers "ads that you run", but "only hides comments made after you create the criteria". Nothing Meta offers touches the comments already under an ad.
- Blocking the commenter barely helps. Of the 534 accounts whose comments were removed under my ads this month, 489 commented exactly once.
- Under my ads, September 1 to 23, 2026: 742 comments, 599 removed. 414 went by a rule with nobody reading them; the largest groups were links (155), emoji strings (110) and pitches judged on meaning (102), and Meta has a word filter for none of them.

To stop spam comments on your ads, switch on the filters Meta runs above the ad, because they reach it: on Facebook the Page's profanity filter, blocked words and Moderation Assist, on Instagram the account's Hidden Words. As of September 23, 2026 those four match words and author signals, and nothing Meta offers reads a comment for what it is or touches the ones already there.

Under my own Facebook and Instagram ads between September 1 and 23, 2026, 742 comments landed and 599 were junk. The four filters would have caught the wordy part. The links, the emoji strings, the phone numbers and the polite pitches, which were most of it, needed a rule of their own.

Below: what lands under an ad and why blocking accounts does not end it, the four layers in Meta's words with what each stops, the shapes they leave, the rules I run instead, the count from my own ads, and the Facebook and Instagram differences.

## What lands under an ad, and why it keeps coming

An ad is a post shown to strangers, and Meta's page on ad comments, read September 17, 2026, starts from that: "For most advertisers, anyone who can see your Facebook, Instagram or Threads ad can react to it or comment on it."

The comments that follow are not conversation. Under my WhatChimp ads the newest two on a September day were a phone-number scam in Bengali and an earn-money pitch in French, with real questions about pricing and the API between them. The six shapes are in [the spam patterns under my own ads](https://commentwatchdog.com/blog/spam-comments-on-facebook-ads-patterns).

The instinct is to block the account. It does little, because the accounts do not come back on their own. Of the 534 accounts whose comments were removed under my ads this month, 489 commented exactly once, and the busiest one commented 11 times. Stopping spam on ads means stopping shapes of comment, not people.

## Layer 1: the Page's word filters reach your Facebook ads

Facebook's controls for a Page carry over to the Page's ads, and Meta says so plainly. Its ad comment best practices, read September 17, 2026: "You can proactively moderate comments by blocking words or applying a profanity filter. When these features are enabled, comments containing these keywords or profanity are hidden. These features are set at the Page level and extend to ads from the Page."

Two switches, both in the Page's settings under Followers and public content:

- **Profanity filter**: Meta's line is "This filter extends to any ads published by your Page."
- **Hide comments containing certain words**: your own list, up to 1,000 entries. Facebook's page on blocked words adds that variations are hidden for you: "If you block the word 'tree', we'll automatically block variations of that word, such as: TREE, t.r.e.e., tr33, treee and #tree."

![The profile's Followers and public content settings, the Hide comments containing certain words row ringed and reading Off.](https://commentwatchdog.com/media/shots/fb-public-content-settings.en.webp)

Both hide, neither deletes, and both need a word to match. The words spam uses under ads are the names of channels and the verbs of a pitch, and a list of them is a list you maintain forever.

## Layer 2: Moderation Assist on the ads you run

Moderation Assist is the automation on a Page or a professional-mode profile, and Meta's note on its scope, read September 5, 2026, is the sentence that matters for advertisers: it "only hides comments made after you create the criteria, and only hides comments on your Page, profile or ads that you run."

So it reaches ads, and it starts from zero. Two of its criteria earn their place under an ad:

- **Spam comments**, "Comment is likely to be spam", Facebook's own classifier with no list to keep
- the four author criteria: no profile picture, no friends or followers, a profile under a week old, a repeat offender

![The Spam comments criterion, Comment is likely to be spam, with Add.](https://commentwatchdog.com/media/shots/fb-ma-spam-criterion.en.webp)

The author criteria are the closest thing Facebook has to stopping the throwaway accounts that comment once. The trade is a real first-time customer with a new profile and no photo, who gets hidden with them.

![Moderation Assist's Choose criteria screen, the four rows about the person commenting ringed.](https://commentwatchdog.com/media/shots/fb-ma-choose-criteria.en.webp)

Facebook also marks some spam by itself, with or without Moderation Assist. Its page on comments marked as spam, read September 10, 2026: "We hide a comment on a post and mark it as spam if: We think it looks suspicious. It's been shared too often in a short amount of time." Under my ads that was the "external" column in the count below, 64 comments in three weeks.

## Layer 3: Hidden Words on your Instagram ads

On Instagram the equivalent is set on the account, and Meta's ad page puts Facebook and Instagram in one sentence: "These filters are set at the Facebook Page, or Instagram and Threads account level, and extend to ads from the Page or account."

Hidden Words has three parts, from Instagram's page read September 5, 2026:

- **Hide unwanted comments**: "Comments containing common offensive words, phrases, or emojis that are likely to be abusive or mean … are automatically hidden. This setting is on by default"
- **Hide spam**, named on the same page beside it, for comments and message requests that may be scams
- **Custom words and phrases**: "words, phrases, numbers, and emojis you want to hide"

On the screen itself, captured on my account on September 8, 2026, the switches carry shorter names: **Hide comments**, **Advanced comment filtering** and **Hide message requests**, with Manage custom words and phrases below them.

![Hidden Words' Hide comments switch is on by default, and Meta says the account-level filters extend to the account's ads.](https://commentwatchdog.com/media/shots/ig-hidden-words-switches.en.webp)

Two things Instagram's own page adds: "hidden comments are still included in your total comment count", and "when a comment or message is hidden, the person who sent it won't know." That is the whole Instagram side; there is no Moderation Assist there, and no delete by rule.

## Layer 4: Ads Manager, one comment at a time

When a comment gets through, Ads Manager is where an advertiser sees it. Meta's steps, read September 17, 2026: the Ads tab, Edit under the ad name, the three dots beside the name, More options, then "Manage Facebook comments" or "Manage Instagram comments". A pop-up shows the ad with its comments, and each one can be hidden or reported.

Two limits sit in that page and its neighbour:

- "You can't manage: comments on Instagram if you don't connect an Instagram profile (ie, you run ads using your Facebook Page identity.)"
- On partnership ads, "Advertisers cannot delete comments on partnership ads." Only the creator can.

This layer is not a way to stop spam. It is where you find out the first three did not, one ad at a time.

## What the four layers leave

The filters match words and author signals. Most of what landed under my ads was neither.

| Shape of comment, under my ads Sep 1 to 23, 2026 | Removed | What Meta offers for it |
|---|---|---|
| A link to a channel, shop or bot | 155 | Moderation Assist's Link in comment, on Facebook only |
| An emoji string, no words | 110 | Hidden Words takes emojis on Instagram; Facebook has no emoji criterion |
| A pitch, a "DM me", a service offer, judged on meaning | 102 | Nothing beyond the "likely to be spam" judgement, which Meta does not define |
| A phone number or an email address | 19 | Nothing; a blocked word cannot match a number |
| A hashtag pile | 15 | Blocked words, if you list the hashtags |
| A blocked word or profanity | 9 | The profanity filter and blocked words |
| An image or a fake support account | 4 | Moderation Assist's Image in comment; nothing for impersonation |

The removed column is what my own rules caught, from my comment log counted September 23, 2026; the right column is my reading of Meta's pages above. And every layer shares one gap in Meta's own words: Moderation Assist "only hides comments made after you create the criteria." The 300 comments already under an old ad stay.

## The rules I run on my ads

CommentWatchDog is an AI comment moderation tool that automatically hides and deletes spam and damaging comments on Facebook and Instagram ads and posts. I built it because the four layers above were on for my own Pages and the spam kept landing. The rules run on every ad and post of a Page at once, and each one hides or deletes on its own. This is the set on my WhatChimp Page, and the one I would start with on any ad account:

1. **URLs**: delete, with your own domain allowed. Every channel, shop and bot link goes, and a link to your site stays.

![The URLs rule on a Facebook Page, on and set to Delete, with the shop's own domain allowed beneath it.](https://commentwatchdog.com/media/shots/moderation-text-phone.en.webp)

2. **Emails and phone numbers**: delete. The WhatsApp numbers and the "send your details" pastes.
3. **Emojis** and **Hashtags**: hide. Cheap to undo if a real fan sends a row of hearts.
4. **Negativity**: hide, with your own sentence about what counts. Under an ad that is the "scam" and "they took my money" pile-on that has nothing to do with your ad.

![The Moderation rules of a Facebook Page, the Negativity switch ringed and on, set to Hide, with the owner's sentence.](https://commentwatchdog.com/media/shots/moderation.en.webp)

5. **Custom topic**: delete, described in plain words. Mine reads "anyone offering a service, a course or a job, or asking to be messaged".
6. **Impersonators**: delete. The fake support accounts that answer your customers before you do.

The [moderation rules guide](https://commentwatchdog.com/help/moderation-rules) has each switch. Then one [Post cleanup](https://commentwatchdog.com/post-cleanup) per old ad clears what was there before the rules were.

## What it stopped under my ads this month

Between September 1 and 23, 2026, my Facebook ads collected 720 comments and my Instagram ads 22. Of the 742, 599 were removed and 143 stayed, most of those questions about pricing, trials and the API.

414 of the removals were a rule acting on its own: links 155, emoji strings 110, negativity 65, my custom topic 37, phone numbers and emails 19, hashtags 15, keywords 5, profanity 4, images 3, one impersonator. 64 were marked outside the app, Facebook's own spam marking among them. 123 I deleted by hand in the inbox, mostly the shapes no rule had yet, and 4 came out in a cleanup of an old ad.

![The inbox of a Facebook Page filtered to comments hidden by the URLs rule: links under ads and posts, each with its Hidden and URL chip.](https://commentwatchdog.com/media/shots/inbox-links-hidden.en.webp)

The number I watch is the 414. Those comments were gone before a buyer scrolled to them, and that is the only meaning of "stop" that matters under an ad: the reader never sees the spam, and the question under it still gets an answer.

## Facebook and Instagram ads: what stops spam on each

- Facebook: the Page's profanity filter and blocked words, Moderation Assist's criteria including the spam judgement and the author signals, Facebook's own spam marking, and Ads Manager's hide and report per comment
- Instagram: Hidden Words' three parts on the account, and Ads Manager only when the Instagram profile is connected to the ad account
- Neither: a rule that reads meaning, a delete by rule, or anything for the comments already under the ad

## Sources

- Meta Business Help Center, Manage comments on your ads in Meta Ads Manager, read September 17, 2026: https://www.facebook.com/business/help/1129470964230971
- Meta Business Help Center, Best practices for comments on Facebook ads, read September 17, 2026: https://www.facebook.com/business/help/297350055160110
- Meta Business Help Center, Manage comments with Moderation Assist on Facebook, read September 5, 2026: https://www.facebook.com/business/help/1753036688579904
- Facebook Help Center, Block words from your Facebook Page, read September 17, 2026: https://www.facebook.com/help/131671940241729
- Facebook Help Center, Approve a comment on your post if it is marked as spam, read September 10, 2026: https://www.facebook.com/help/134035947160858
- Instagram Help Center, Hide comments or message requests you don't want to see on Instagram, read September 5, 2026: https://help.instagram.com/700284123459336
- Meta Business Help Center, Manage comments on your Facebook and Instagram partnership ads, read September 17, 2026: https://www.facebook.com/business/help/725874135641929
- My own comment log, three Facebook Pages and one Instagram account, comments under ads published September 1 to 23, 2026, counted September 23, 2026.

Common questions:
Question: Can I just turn off comments on my ads?
Answer: Some accounts and ad formats offer it and most do not, as of September 2026, and the switch takes the questions and the social proof with the spam. The filters below keep the comments and remove the junk.
Question: Do Facebook's word filters work on ads?
Answer: Yes. Meta's page on ad comments, read September 17, 2026, says the profanity filter and blocked words "are set at the Page level and extend to ads from the Page", and that Hidden Words does the same for Instagram and Threads.
Question: Why does blocking the spammer not stop the spam?
Answer: Because the next comment comes from a different account. Under my own ads between September 1 and 23, 2026, 489 of the 534 accounts whose comments were removed commented exactly once.
Question: What stops the spam Meta's filters miss?
Answer: A rule that reads the comment rather than matching a word: links, phone numbers, emoji strings, pitches and negativity, on every ad at once, plus a cleanup of what was already there. CommentWatchDog is an AI comment moderation tool that automatically hides and deletes spam and damaging comments on Facebook and Instagram ads and posts. 7 days free, no card.

Plain text of this post: https://commentwatchdog.com/blog/stop-spam-comments-on-ads.md
All posts for assistants: https://commentwatchdog.com/blog/llms.txt