Community Comment Reply Prompt

What this prompt does

A community comment reply prompt turns a pasted batch of social media comments into individual reply drafts that all sound like the same brand — questions answered in the first sentence, compliments thanked for the specific thing they praised, and the risky ones flagged for a human instead of drafted. It's built for the part of social media work nobody posts about: the comment section itself.

Comments don't arrive one at a time. A post that performs brings forty replies in an afternoon, and somewhere around number thirty, whoever's answering them starts writing on autopilot. The first reply of the day and the last one shouldn't sound like two different companies, but by 5 p.m. they usually do.

This is a different job from review responses. The customer review response prompt writes for the permanent record — a Google or Yelp review sits there for years, star-rated, read by next year's customers, so each response is a small essay. Comments move faster and die sooner. They're conversational, they're public but ephemeral, and there are twenty of them before lunch. This prompt handles that volume without letting the voice drift.

The unglamorous work pays, and there's a number on it. A 2018 Harvard Business Review study by Wayne Huang and colleagues, drawn from Twitter customer service data, found airline customers who got a reply to their tweet were willing to pay about $9 more for a future ticket — and the single biggest lift came from replying to positive comments, worth around $28. Most brands only answer complaints. The data says the thank-yous are worth more.

The prompt

Draft replies to the social media comments below. Reply to each one individually, numbered to match my list.

**Platform:** [INSTAGRAM / TIKTOK / YOUTUBE / LINKEDIN / FACEBOOK / X]
**Who we are:** [ONE LINE — e.g., "three-person coffee gear brand, playful but not jokey"]
**Voice:** [2–3 ADJECTIVES, PLUS PASTE ONE REAL REPLY YOU'VE SENT THAT SOUNDED RIGHT]
**Context right now:** [CURRENT CAMPAIGN, KNOWN ISSUES, PRODUCT FACTS COMMENTERS ARE ASKING ABOUT]
**Never say:** [BANNED PHRASES, CLAIMS WE CAN'T MAKE, COMPETITOR NAMES]

**The comments:**
[PASTE AS A NUMBERED LIST]

Rules:
- Match each commenter's energy. Answer questions in the first sentence. Thank compliments for the specific thing praised. A joke can get a light reply back.
- Keep replies under 30 words unless a question genuinely needs steps.
- Vary the openings across the batch. If three replies start the same way, rewrite two of them.
- If a reply needs a fact I didn't give you — order status, restock dates, pricing — write [VERIFY] in its place instead of guessing.
- Do NOT draft a reply for: legal threats, safety complaints, harassment, messages from journalists, or a complaint that looks like it could escalate. For those, output [ESCALATE] plus one line on why.
- No corporate filler. Never "Thanks for reaching out," never "We appreciate your feedback."

How to use it

1. Batch by platform, not by day. Tone that lands on LinkedIn dies on TikTok, so run one batch per platform. And run them often — in the same HBR study, a reply within five minutes was worth about $20 to airline customers, falling to roughly $2 after an hour. Three short sweeps beat one nightly cleanup.
2. Paste a real reply, not just adjectives. "Friendly but dry" means something different to every model; one actual reply you were happy with pins the register better than any description. If you keep re-explaining your voice in every prompt, the brand voice prompt builds the reusable version once.
3. Read the flags before the drafts. The [ESCALATE] lines are the real output. A batch that comes back clean is nice; the one comment correctly routed to a human before it became a quote in someone's screenshot thread is the reason the rule exists.
4. Sign the ones that matter. The same study found a reply signed with a name or initials was worth $14 more than an unsigned one. Adding "— Priya" costs three characters.

Example output

Inputs: a three-person coffee gear brand on Instagram, the morning after a grinder restock post. Context supplied: "V2 grinder restocked yesterday, grinds espresso through French press, coarseness guide linked in bio, standard shipping is 3–5 business days."

The comments:

1. "been waiting 3 months for this restock 😭 just ordered"
2. "does it handle french press or just espresso?"
3. "ordered two weeks ago and still nothing. hello???"
4. "my wallet hates this account"
5. "the burrs chipped after a week and I found metal flakes in my grounds"

1. Three months of waiting deserves a good first brew — glad you got one before they went again.

2. French press works — it grinds espresso through coarse. The coarseness guide is linked in bio.

3. Two weeks is longer than our 3–5 day standard, and that's on us. DM us your order number and we'll chase it today — [VERIFY: no order lookup available to me].

4. Your wallet started it by following a coffee account.

5. [ESCALATE] Safety complaint — metal fragments in a food product needs a human response and probably your QA process, not a comment reply.

Five comments, five different openings, and the two flags land where they should: reply 3 gets drafted but refuses to invent an order status, and reply 5 doesn't get drafted at all. That last distinction is the whole design — a shipping gripe is a conversation, metal in coffee grounds is an incident.

Variations

For DMs

"These are private messages, not public comments. No audience is watching, so drop the performance — answer each person directly, use their name if they gave one, and where a public reply would say 'DM us,' actually resolve it or say exactly what happens next."

For creator accounts

"This is a personal creator account, not a brand. Write in first person as me. The commenters are viewers, not customers — warmer, more casual, and it's fine to leave a question open or just react."

After a public mistake

When something went wrong and the comments all say so, don't generate forty different apologies — pick one honest holding line, pin it, and repeat it with small variations. Add to the prompt: "We made a mistake — [WHAT HAPPENED]. Every reply should acknowledge it plainly, point to [WHERE THE FULL RESPONSE LIVES], and never get defensive." The apology email prompt writes the direct message to the people actually affected; the comment section just needs to see you standing still.

Common pitfalls

• Don't send AI drafts under grief or serious personal disclosures. When someone comments that they bought your product for a parent who's since died, that reply gets written by a person, even though a model could fake it. Faking it is the problem.
• Don't use it to argue with trolls. The prompt matches energy, and a troll's energy is the one thing you never want matched. Hide, ignore, or block — a well-drafted comeback is still feeding.
• Don't run a 200-comment mega-batch. Voice drifts across a long generation the same way tired humans drift. Twenty to thirty comments per run keeps the replies sharp enough to spot-check.
• Try instead: post one reply from each batch yourself, unedited by the model — an inside joke, a genuine reaction. It keeps your own register alive in the section, which is what step 2 feeds back into tomorrow's batch.

Who uses this prompt

• Social media managers: the after-lunch comment sweep across client accounts, where the client's voice has to survive being the fourth brand you've written for today. More in prompts for social media managers.
• Small business owners: answering their own comment section at 9 p.m. after doing every other job in the business since 7 a.m.
• Content creators: the day a video takes off and the comment count grows faster than any one person types.

ChatGPT, Claude, and Gemini all run this prompt unmodified. The field doing the quiet heavy lifting is the pasted example reply — Anthropic's prompt engineering guidance ranks showing the model an example among the most effective techniques there is, and one real reply outperforms a paragraph describing your tone. Which mirrors how comment sections actually work: nobody screenshots the polished reply. They screenshot the one where a brand answered a joke with a better one.