Replying to comments on your own posts is the job of answering the people who took the time to respond to you, and it is a Skill job you can finish in one sitting of twenty to forty minutes. Give the AI the post, the comments word for word and a few replies you wrote yourself, and it returns one reply draft per comment that answers the person and gives them a reason to say something back. This page covers the comments under your own posts on any channel. Answering questions in other people's groups lives on the Facebook page, and finding conversations about you elsewhere lives on social listening. The AI drafts. You read, edit and post.
- Shape
- Skill
- Sitting
- one sitting
The shape: Skill
This job is a Skill because it is one repeatable move: a comment goes in, a reply in your voice comes out. There is nothing to connect, no tool to install and no chain where one output feeds the next. You do it the same way every time you sit down with your notifications.
What makes it a skill and not a one-off prompt is the brief you keep. The first replies will sound too cheerful, too long or too much like a help desk. You tell the AI what you cut and why, and the next round starts closer to how you talk to customers. Keep the brief in a note, and every sitting starts from the better version.
The job is small on purpose. Most comments on a small business page are short: a heart, a "love this", a question about price or hours, a tag of a friend. Each one is a person who stopped scrolling, and the reply is how they learn that a real person runs the page. A thread with replies also tells the next reader that the business is open and paying attention.
What you give the AI
Gather these before you open the chat. Every item is something only you have, which is why the replies end up sounding like you.
- The post itself. The caption, plus one line describing the photo or video, since the AI cannot always see what commenters are reacting to.
- Every comment, word for word, with the commenter's first name. Copy them in the order they appear. Note any that are replies to other comments, so the AI knows who is talking to whom.
- What you know about each commenter, in a few words. "Regular customer since spring", "never heard of her", "my cousin". Leave it blank when you do not know.
- Three replies you wrote yourself and liked. From any channel, or texts to customers. These set the voice better than any description of it.
- The facts the AI may use. Hours, booking link, service area, what you do and do not offer, anything current like a holiday closure. If a commenter asks something outside this list, the AI should flag it, not guess.
- Your do-not-say list. Prices you do not post in public, promises about results, anything your license or trade rules out, and anything you would rather move to a private message.
- Which comments to skip. Spam, bots and anything abusive. Hide or delete those yourself; they do not get a reply.
What comes back
Paste the brief below into the AI you already use and fill in the brackets, except [CHECK], which the AI uses to mark facts it needs from you. Run it once per post, in the same conversation, so it keeps your voice and your corrections from one post to the next.
You are helping a small business owner reply to comments on their own social media post.
The owner will read, edit and post every reply from their own account.
The business: [what you do, where, for whom]
The channel: [Facebook / Instagram / LinkedIn / other]
How the owner writes (match this voice): [paste 3 replies or texts you wrote]
Facts you may use: [hours, booking link, service area, current news]
Do not say: [your list]
THE POST
Caption: [paste it]
What the photo or video shows: [one line]
THE COMMENTS
[Name] ([what you know about them, or "unknown"]): [comment, word for word]
[Name] ([...]): [comment]
[repeat for each comment]
Write one reply per comment.
Each reply:
- Uses the commenter's first name once, or not at all if that reads stiff.
- Answers or responds to what they actually said in the first sentence.
- Adds one small thing that gives them a reason to reply: a short question about their situation, a detail they did not know, or a thank-you that names what they noticed.
- Is one to three sentences, plain words, no hashtags, no emoji unless the owner's own replies use them.
- Never repeats the same opening as another reply on this post.
For a question you cannot answer from the facts above, write the reply with [CHECK] where the fact goes.
For a complaint, write one calm public reply that acknowledges the specific problem and offers a private way to sort it out, then stop.
Do not invent prices, times, results, reviews or customers.
Under the replies, list in one line each:
- any comment that looks like spam, a bot or bait, and should get no reply
- any comment that should move to a private message or be hidden, such as one that shares personal details in public, and whyWhat comes back is a short reply under each comment, in the order you pasted them, followed by two short lists. The lists matter as much as the replies. If the AI flags that a comment is a giveaway bot or that a customer just shared their home address, the right move is to hide that comment, not answer it.
When the sitting is done, paste back the replies you actually posted, ask the AI what you changed and why, and ask it to fold those changes into the brief. Save the updated brief. That is how the skill gets better each week.
How to judge it
Read each reply as the person who wrote the comment. You can reject a bad one in a few seconds if you know what to look for.
It could go under any comment. "Thanks so much for the support" fits every comment ever written, which is why it tells the commenter nobody read theirs. A good reply could only go under that one comment.
Every reply starts the same way. Five replies that all open with "Thank you" or the commenter's name followed by an exclamation read like a form letter. Vary the openings, and let some replies skip the thanks entirely and go straight to the answer.
The question at the end is fishing. "What's your favorite season for a patio?" bolted onto a reply about a broken gate is engagement bait, and people can tell. A good follow-up question is one you would actually want answered.
It is longer than the comment deserves. A heart emoji does not need three sentences back. Match the length to the comment, and let some replies be four words.
It states a fact you did not give it. An AI filling a gap will name a price, a wait time or an opening date in a confident tone. Delete every fact you did not supply, and fill every [CHECK] yourself or move that question to a private message.
It argues in public. A reply to a complaint that corrects the customer's version of events, even politely, makes the thread worse for everyone reading it. One acknowledgment and one private next step is the whole reply.
What stays with you
The AI writes drafts. These parts of the job stay with you, and they are the parts commenters feel.
Knowing who is talking. Only you know that the person asking about hours is a regular, that the friendly stranger is a competitor, or that the tagged friend is a likely customer. That knowledge changes the reply, and it lives in your head, not in the comment.
Every fact. Hours, prices, availability and anything a customer might hold you to are yours to check before you post. In health, law, finance, real estate or a licensed trade, decide ahead of time what you never discuss in public comments, and move those questions to a private message every time.
Complaints and anything sensitive. The first public reply can be drafted. The conversation after it, by phone or message, is yours to have, and it is where an unhappy customer either stays or leaves.
Timing and the post button. A reply a day later still lands; a reply three weeks later reads as an afterthought. You post each reply from your own account, and you decide when a thread has said enough. Software that writes and publishes your posts does not change this. Every Boomp plan leaves replies to comments with you.
An editorial example
This is an invented example to show the skill, not a real business.
Say you run a small landscaping company with two crews. On Tuesday you posted a photo of a finished backyard with a new paver patio and some native plants. By Thursday it has six comments: two hearts, a neighbor of the client saying "saw this going up, looks great", a stranger asking "do you do just the plants or only full yards", a friend tagging her sister, and an account with no photo offering to "boost your page".
You paste the post, the six comments, three replies you wrote last month and your facts: your service area, that you take small planting-only jobs from October to March, and your booking link.
The AI returns five replies and flags the boost offer as spam. The reply to the neighbor thanks her and mentions that the native plants fill in by next summer, which she can watch from her side of the fence. The reply to the stranger answers plainly that yes, you take planting-only jobs in the fall and winter, and asks what she has in mind for her yard. One draft for the tagged sister promises "a free estimate this week", which you never offer, so you cut it to a short hello and the booking link. The heart replies come back as one sentence each, and you shorten one to just "Thank you, Dana." You hide the spam comment, post the five replies and close the app in under half an hour.
