Guide: Reply to comments on your ads and hide the spam
This page: https://commentwatchdog.com/help/ai-agents/reply-to-comments-on-your-ads-and-hide-the-spam
Topic: AI agents
Time: about 4 minutes
Who can do it: Owner, Administrator or Manager
Author: https://anantgupta.xyz (Anant Gupta, founder of CommentWatchDog)
Start in the app: https://app.commentwatchdog.com/en/settings/agents
Result when done: Real questions under your ads get an answer as your page, and the spam is gone before anyone reads it, both without you.
Context: Under an ad, two things happen at once: strangers ask real questions, and spam lands between them. The moderation rules hide or delete the spam the moment it arrives, and an AI agent answers the real questions as your page, from the facts you gave it. Both run on comments under ads and under posts, on Facebook and on Instagram. With the recommended audience the agent replies only to visible comments that need a reply, so it never answers a comment the rules hid. Set the rules first, then the agent, and let it draft for a week before it sends.

Before you start:
1. Rules first: the agent answers what is left after the rules have hidden or deleted the spam.
2. The agent replies as the page, and each reply is credited to the agent in the app.
3. Start on Auto-drafts and read a week of drafts before Auto-replies.

Steps:
Step 1 of 4: Hide the spam first: rules on, applied to ads.
Where: Left sidebar, Pages, the page, then the Moderation tab. Switch on the rules, and under the rules set Apply moderation to, to Both posts and ads or Ads only. From then on a scam link, a fake support account or an insult under the ad is hidden or deleted before anyone reads it.
Link: https://app.commentwatchdog.com/en/inbox (Open the app)
Screenshot: https://commentwatchdog.com/media/shots/moderation-whitelist.en.webp (Apply moderation to: both, ads only, or posts only.)
Screenshot: https://commentwatchdog.com/media/shots/moderation-whitelist-phone.en.webp (The same on a phone.)
You should see: Spam under the ad arrives hidden or deleted, with the rule named on the row.

Step 2 of 4: Create the agent for the page and give it the facts.
Where: Settings, AI agents, Create agent. Name it, tick the page the ads run from, write the persona, then paste your shipping, returns, prices and sizes into its knowledge and add FAQ pairs for the questions you answer every day.
Link: https://app.commentwatchdog.com/en/settings/agents (Open AI agents)
Screenshot: https://commentwatchdog.com/media/shots/agent-detail.en.webp (Pages this agent covers.)
Screenshot: https://commentwatchdog.com/media/shots/agent-phone.en.webp (The same on a phone.)
You should see: An agent assigned to the page, with its persona and its knowledge saved.

Step 3 of 4: Set who gets a reply: visible comments that need one, under ads.
Where: On the agent, Reply to. Visible comments that need a reply is the recommended choice: the agent judges which comments are real questions and leaves praise, spam and hidden comments alone. Apply to sets whether it answers under ads, posts, or both. Add the phrases you never want answered under Do not reply to.
Screenshot: https://commentwatchdog.com/media/shots/agent-mode.en.webp (Reply to, on the agent.)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode-phone.en.webp (The same on a phone.)
You should see: The agent labels the comments it would answer with Needs reply, and nothing else.

Step 4 of 4: Start on Auto-drafts, then switch to Auto-replies.
Where: Agent mode, on the agent. Auto-drafts writes a reply for every matching comment and waits in the composer for you to send. When a week of drafts stops surprising you, choose Auto-replies and it sends on its own, as the page.
Screenshot: https://commentwatchdog.com/media/shots/agent-mode.en.webp (Agent mode: Auto-drafts, then Auto-replies.)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode-phone.en.webp (The same on a phone.)
You should see: Replies go out under the ad on their own, credited to the agent, and the spam stays hidden.

If something is off:
Problem: It replied to a spam comment
Fix: The rules missed it, so it was visible. Describe that spam under Custom topics so the rules hide it, and add its wording to Do not reply to on the agent.
Problem: Nothing was answered under my ad
Fix: Check that the page is ticked on the agent, that Apply to includes ads, and that Agent mode is not Off. Off only labels and powers Generate.
Problem: It answered praise
Fix: Narrow Reply to, to visible comments that need a reply, and add the case to Do not reply to.
Problem: A reply failed to send
Fix: Meta refused it or is limiting actions; the row keeps the words so you can send them yourself.

Common questions:
Question: Can an AI reply to comments on my Facebook and Instagram ads automatically?
Answer: Yes. An AI agent in CommentWatchDog answers comments under your ads and posts as your page, from the persona, knowledge and FAQ pairs you gave it, on Facebook and on Instagram. Apply to sets whether it answers under ads, posts, or both, and Agent mode sets whether it drafts for you or sends on its own.
Question: Will the agent reply to spam?
Answer: Not with the recommended audience. The moderation rules hide or delete the spam as it arrives, and an agent set to Visible comments that need a reply answers only visible comments it judges to be real questions. Anything you add to Do not reply to is left alone whatever else matches.
Question: Does an AI agent cost extra?
Answer: No. CommentWatchDog counts one credit per incoming comment on live moderation, and replies never count, whether a person, an automation or the agent wrote them. AI agents start on Starter.
Question: Can I see the replies before they go out?
Answer: Yes. On Auto-drafts the agent writes a reply for every matching comment and waits in the composer; you read it and send. Auto-replies sends on its own, and Pause on the agent list stops it at any time without losing its knowledge.
Question: Who does the reply appear to come from?
Answer: From your page, like any reply you send yourself. Inside CommentWatchDog the reply is credited to the agent, and Insights counts replies by user, automation and agent separately.

Related guides:
- Draft mode or auto mode: https://commentwatchdog.com/help/ai-agents/draft-mode-or-auto-mode
- Who gets a reply, and who never does: https://commentwatchdog.com/help/ai-agents/who-gets-a-reply-and-who-never-does
- Ads only, posts only, or both: https://commentwatchdog.com/help/moderation-rules/ads-only-posts-only-or-both

Plain text of this guide: https://commentwatchdog.com/help/ai-agents/reply-to-comments-on-your-ads-and-hide-the-spam.md
All guides for assistants: https://commentwatchdog.com/help/llms.txt