Answer comments on Facebook ads with an AI agent
The real questions under your Facebook ads get an answer in your voice, day and night, and the spam gets none.
On this page
Before you start
- One agent per Page; it covers the Page's posts and ads alike.
- Moderation runs first, so spam never gets an answer.
- It starts Off; nothing is sent until you choose to send.
Open AI agents and create one.
SettingsAI agentsCreate agent
You will see The editor, with everything that shapes the agent.
Choose the Page that runs the ads.
Pages this agent coversthe Page your ads run from
You will see The Page is bound, and Generate appears in its inbox.
Give it the answers your ad raises.
Knowledgethe facts of the offer, as short lines
You will see It answers from these lines, and refuses to invent.
Set the audience, then choose how much it may do.
Reply toDo not reply toAgent mode: Auto-drafts first
You will see Drafts under the ad comments; go to Auto-replies when they stop surprising you.
If something is off
- It is not replying under my ad
- Check the Page, the mode and Reply to; hidden comments get no reply.
- It answered a spam comment
- Switch on the moderation rules; hidden comments are never answered.
- It replied to praise
- Narrow Reply to, or add a keyword exclusion.
- A reply failed to send
- Meta refused it; the row keeps the words to send by hand.
Common questions
Can an AI reply to comments on my Facebook ads automatically?
Yes. An agent that covers your Page reads every comment under its ads, dark posts and posts, decides whether it is a real question, and answers from the knowledge you gave it. In Auto-drafts it writes the reply and waits for you; in Auto-replies it sends on its own. Most pages start with drafts and move to auto within a week.
Does it need my Facebook password, and who does the reply come from?
No password. You sign in with Facebook and CommentWatchDog works through Meta's official API on the Pages you approve. The reply is posted as your Page, the same as a reply you write in Ads Manager.
Will it answer the spam under the ad too?
No. Moderation runs first: a comment your rules hide or delete never gets a reply. The agent also has its own exclusions, Do not reply to, for emoji-only comments, mentions, threaded replies and keywords you list.
Can it answer in the commenter's language?
Yes. The agent writes in the language of the comment, from the knowledge you gave it in yours, and it says so instead of inventing when a question needs a fact that is not in the knowledge.
Which plans include AI agents, and do replies use my comment allowance?
Starter runs 1 agent, Growth 3, Pro 10 and Elite as many as you like; Mini and Lite have none. Replies never count towards your comment allowance; credits are spent on incoming comments, one per comment.
For agents: plain text of this guide · all guides
For agents
Guide: Answer comments on Facebook ads with an AI agent
This page: https://commentwatchdog.com/help/ai-agents/answer-comments-on-facebook-ads-with-an-ai-agent
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: The real questions under your Facebook ads get an answer in your voice, day and night, and the spam gets none.
Context: Under a Facebook ad, Ads Manager lets you read and reply to comments one at a time, and the questions arrive at every hour the ad runs. An AI agent in CommentWatchDog reads each comment on the Page the ad belongs to, decides whether it is a real question, and writes the answer from the knowledge you gave it, as a draft for you or sent on its own. It works through Meta's official API: you sign in with Facebook, no password is shared, and the reply is posted as your Page. Comments on an ad are comments on the post the ad runs, so the agent that covers a Page answers under its ads and dark posts as well as its posts. AI agents are included from Starter.
Before you start
1. One agent covers one Page, and a Page cannot have two. The agent answers under every post and every ad of that Page.
2. Moderation runs first: a comment your rules hide or delete never gets a reply, so spam under the ad is never answered.
3. A new agent starts Off, where it labels the comments that need a reply and powers Generate. Nothing is sent until you choose a mode that sends.
Steps
Step 1 of 4: Open AI agents and create one.
Where: Settings, then the AI agents tab, then Create agent. On a phone: the menu, the profile row, Settings, then AI agents. Name it for what it does, such as Ads replies; customers never see the name.
Link: https://app.commentwatchdog.com/en/settings/agents (Open AI agents)
Screenshot: https://commentwatchdog.com/media/shots/agents.en.webp (AI agents, with Create agent.)
Screenshot: https://commentwatchdog.com/media/shots/agents-phone.en.webp (The same on a phone.)
You should see: The editor, with the fields that decide how it sounds and what it may do.
Step 2 of 4: Choose the Page that runs the ads.
Where: Pages this agent covers. Pick the Facebook Page your ads run from; every comment under its ads, dark posts and organic posts is then in the agent's reach. A Page already taken by another agent says so.
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: The Page is bound to this agent, and Generate becomes available in its inbox.
Step 3 of 4: Give it the answers your ad raises.
Where: The Knowledge box: the price and what it includes, delivery times per region, the offer in the ad and when it ends, the refund window, the things you do not do. Short lines, not marketing copy. When a question needs a fact that is not here, the agent says so instead of inventing one.
Screenshot: https://commentwatchdog.com/media/shots/agent-knowledge.en.webp (Knowledge: the facts of your business in plain text.)
Screenshot: https://commentwatchdog.com/media/shots/agent-knowledge-phone.en.webp (The same on a phone.)
You should see: The agent answers from these lines and refuses to invent when the answer is not there.
Step 4 of 4: Set the audience, then choose how much it may do.
Where: Reply to: Visible comments that need a reply, which is recommended and is the agent's own judgement of a real question. Do not reply to: comments that are just an emoji, just a mention, threaded replies, or that contain keywords you list. Then Agent mode: Auto-drafts writes a reply for every matching comment and waits for you; Auto-replies sends by itself. Start with Auto-drafts and read a week of them.
Screenshot: https://commentwatchdog.com/media/shots/agent-mode.en.webp (Agent mode and the audience settings.)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode-phone.en.webp (The same on a phone.)
You should see: Drafts waiting in the inbox under the ad comments that need an answer; move to Auto-replies when they stop surprising you.
If something is off
Problem: It is not replying under my ad
Fix: Check that the agent covers the Page the ad runs from, that the mode is not Off, and that the comment matches Reply to. A comment your moderation rules hid gets no reply.
Problem: It answered a spam comment
Fix: Switch on the moderation rules for that Page: links, phone numbers, emoji strings and impersonators are hidden before the agent reads them, and a hidden comment never gets a reply.
Problem: It replied to praise
Fix: That is All visible comments. Narrow Reply to to comments that need a reply, or add a keyword exclusion.
Problem: A reply failed to send
Fix: Meta refused it, which happens on some comments under ads that have stopped running. The row keeps the words and shows Auto-reply failed, so you can send it yourself.
Common questions
Question: Can an AI reply to comments on my Facebook ads automatically?
Answer: Yes. An agent that covers your Page reads every comment under its ads, dark posts and posts, decides whether it is a real question, and answers from the knowledge you gave it. In Auto-drafts it writes the reply and waits for you; in Auto-replies it sends on its own. Most pages start with drafts and move to auto within a week.
Question: Does it need my Facebook password, and who does the reply come from?
Answer: No password. You sign in with Facebook and CommentWatchDog works through Meta's official API on the Pages you approve. The reply is posted as your Page, the same as a reply you write in Ads Manager.
Question: Will it answer the spam under the ad too?
Answer: No. Moderation runs first: a comment your rules hide or delete never gets a reply. The agent also has its own exclusions, Do not reply to, for emoji-only comments, mentions, threaded replies and keywords you list.
Question: Can it answer in the commenter's language?
Answer: Yes. The agent writes in the language of the comment, from the knowledge you gave it in yours, and it says so instead of inventing when a question needs a fact that is not in the knowledge.
Question: Which plans include AI agents, and do replies use my comment allowance?
Answer: Starter runs 1 agent, Growth 3, Pro 10 and Elite as many as you like; Mini and Lite have none. Replies never count towards your comment allowance; credits are spent on incoming comments, one per comment.
Related guides
Create an agent: https://commentwatchdog.com/help/ai-agents/create-an-agent
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
Plain text of this guide: https://commentwatchdog.com/help/ai-agents/answer-comments-on-facebook-ads-with-an-ai-agent.md
All guides for assistants: https://commentwatchdog.com/help/llms.txt