Who gets a reply, and who never does
The agent answers the people you want answered and stays out of everything else.
On this page
Before you start
- The recommended scope is visible comments that need a reply.
- Including hidden comments is rarely what you want.
- Do not reply to wins over everything else.
Set Reply to.
Reply to, on the agent
You will see A narrower scope means fewer surprises.
Set Do not reply to.
Do not reply to, on the agent
You will see Those comments are left alone regardless.
Add the keywords you never want answered.
Specific keywords, inside Do not reply to
You will see Those words are never answered by the agent.
If something is off
- It is answering praise
- Narrow the scope, or exclude a keyword.
- It missed a real question
- Change the intent prompt that decides what needs a reply.
- It replied to a hidden comment
- Change the scope to visible comments only.
- It replied under its own reply
- Add threaded replies to Do not reply to.
Common questions
Which Reply to setting should I choose for an AI agent?
Visible comments that need a reply, which is the recommended scope and the agent's own judgement of a real question. The other options are All visible comments and All comments including hidden ones.
What happens if a comment matches Reply to but also Do not reply to?
It is left alone. Do not reply to wins over everything else, even when the comment matches the audience scope.
What kinds of comments can I stop the agent from answering?
Comments that are only an emoji, only a mention, threaded replies, or that contain specific keywords you list under Do not reply to. Legal words, competitor names and anything that should reach a human are the usual entries.
For agents: plain text of this guide · all guides
For agents
Guide: Who gets a reply, and who never does
This page: https://commentwatchdog.com/help/ai-agents/who-gets-a-reply-and-who-never-does
Topic: AI agents
Time: about 3 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 agent answers the people you want answered and stays out of everything else.
Context: Two settings draw the line. Reply to says which comments are in scope at all. Do not reply to names the kinds you never want answered, whatever else matches.
Before you start
1. The recommended scope is visible comments that need a reply, which is the agent own judgement of a real question.
2. Hidden comments can be included, which is almost never what you want in auto mode.
3. Do not reply to wins over everything else.
Steps
Step 1 of 3: Set Reply to.
Where: On the agent: Visible comments that need a reply, which is recommended; All visible comments; or All comments including hidden ones.
Link: https://app.commentwatchdog.com/en/settings/agents (Open AI agents)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode.en.webp (Reply to.)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode-phone.en.webp (The same on a phone.)
You should see: The narrower the scope, the fewer surprises when you move to auto.
Step 2 of 3: Set Do not reply to.
Where: The exclusions: comments that are just an emoji, just a mention, threaded replies, or that contain specific keywords you list. These are the comments where a reply adds nothing or makes things worse.
Screenshot: https://commentwatchdog.com/media/shots/agent-mode.en.webp (Do not reply to.)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode-phone.en.webp (The same on a phone.)
You should see: Those comments are left alone even when everything else matches.
Step 3 of 3: Add the keywords you never want answered.
Where: Specific keywords, inside Do not reply to. Legal words, competitor names, and anything that should reach a human are the usual entries.
Screenshot: https://commentwatchdog.com/media/shots/agent-mode.en.webp (The words under Do not reply to.)
Screenshot: https://commentwatchdog.com/media/shots/agent-mode-phone.en.webp (The same on a phone.)
You should see: A comment containing one of those words is never answered by the agent.
If something is off
Problem: It is answering praise
Fix: That is All visible comments. Narrow it to comments that need a reply, or add a keyword exclusion.
Problem: It missed a real question
Fix: The agent judged it as not needing a reply. The intent prompt is what decides that, and you can change it.
Problem: It replied to a hidden comment
Fix: The scope includes hidden comments. Change it to visible only.
Problem: It replied under its own reply
Fix: Add threaded replies to Do not reply to.
Common questions
Question: Which Reply to setting should I choose for an AI agent?
Answer: Visible comments that need a reply, which is the recommended scope and the agent's own judgement of a real question. The other options are All visible comments and All comments including hidden ones.
Question: What happens if a comment matches Reply to but also Do not reply to?
Answer: It is left alone. Do not reply to wins over everything else, even when the comment matches the audience scope.
Question: What kinds of comments can I stop the agent from answering?
Answer: Comments that are only an emoji, only a mention, threaded replies, or that contain specific keywords you list under Do not reply to. Legal words, competitor names and anything that should reach a human are the usual entries.
Related guides
How the agent decides Needs reply: https://commentwatchdog.com/help/ai-agents/how-the-agent-decides-needs-reply
Draft mode or auto mode: https://commentwatchdog.com/help/ai-agents/draft-mode-or-auto-mode
Tone, delay, mention and archive: https://commentwatchdog.com/help/ai-agents/tone-delay-mention-and-archive
Plain text of this guide: https://commentwatchdog.com/help/ai-agents/who-gets-a-reply-and-who-never-does.md
All guides for assistants: https://commentwatchdog.com/help/llms.txt