The Social Media Reply System

Turn a busy comment queue into clear decisions: reply, move to a private support channel, escalate, moderate, or leave it alone.

For: founders, creators, community managers, and small teams using AI to draft replies without giving up judgment or brand control.

You will finish with: a reply brief, a safe triage rule, 6 platform-specific playbooks, and a copy-ready prompt for drafting replies.

AI drafts. A person decides.

A social reply can affect a customer, a public record, and the people watching the conversation. Use AI to reduce drafting time, not to automate judgment.

Before sending any comment to an AI tool:

• Remove email addresses, phone numbers, order numbers, and other personal data.
• Keep the original post and the specific comment together so the draft has context.
• Identify the platform, audience, desired outcome, and brand boundary.
• Mark any factual statement that requires verification.
• Keep a human reviewer in control of posting, hiding, reporting, or escalating.

Platform policies change. The current official rules for X, Meta platforms, TikTok, LinkedIn, and Reddit should be checked before formalizing a moderation process.

Step 1: triage before you draft

Use the comment's risk and intent to choose the next action.

The decision to avoid replying is valid. A reply is useful only when it answers, clarifies, acknowledges, or protects.

Step 2: build the context packet

AI drafts become generic when the input is generic. Give the model a compact context packet instead of a bare comment.

Platform:
Original post summary:
Comment, with personal data removed:
Comment intent: question, praise, disagreement, support issue, correction, provocation, or risk
Audience watching the thread:
Desired outcome:
Brand voice:
Facts the reply may use:
Facts that still need verification:
Words, claims, or offers to avoid:
Maximum length:
Preferred next step:

Keep the factual packet short. If the reply depends on a current claim, attach the official source or remove the claim.

Step 3: use the three-part reply shape

Most useful replies contain 3 moves:

1. Recognize: show that you understood the actual point.
2. Contribute: add an answer, clarification, evidence, or practical next step.
3. Continue or close: invite a useful response, route support privately, or end cleanly.

Not every reply needs a question. Asking a vague question only to increase engagement can feel mechanical. Close the exchange when the issue has been resolved.

X: clarify one point in public

X rewards fast public conversation, but speed makes context collapse more likely. Keep the reply focused on one point and avoid turning a small disagreement into a long defensive thread. X prohibits targeted harassment, private-information exposure, and artificial platform manipulation. Official X rules

Scenario: skeptical reply to a product claim

Original post: "A good prompt can make a weekly report easier to draft."

Comment: "That sounds like another excuse to publish made-up numbers."

Weak draft: "AI is the future, and people who learn it will stay ahead."

Platform-native reply: "That risk is real if the model is asked to invent the analysis. Our workflow starts with an approved spreadsheet, asks for the calculation, and checks one key number manually before anything is published."

Why it fits X: it addresses the criticism directly, contributes a verifiable process, and does not demand further engagement.

Check before posting: can the account prove that this is its actual workflow? If not, rewrite it as a recommendation rather than a statement of practice.

Instagram: connect the reply to the visual

Instagram comments often refer to something visible in the post or reel. Name the element the person reacted to so the response does not feel copied. For a customer-specific issue, acknowledge the problem publicly and move account details to an approved support channel. Meta's Community Standards apply across its platforms and should guide moderation decisions. Official Meta standards

Scenario: a practical question under a carousel

Post: a carousel showing a 3-step research checklist.

Comment: "Do I really need to open every source?"

Weak draft: "Great question! Thanks for checking out our content."

Platform-native reply: "For an important claim, yes. The source slide is a discovery step, not proof by itself. Open the cited page and confirm that it supports the exact sentence you plan to use."

Why it fits Instagram: it points back to the carousel, gives a concrete answer, and is short enough to scan in the comment thread.

Check before posting: avoid asking the person to share private details in a public comment. If support is needed, name the official private route without requesting an order number in public.

Facebook: answer for the wider thread

Facebook comments can remain visible to a broad audience long after the first exchange. Write an answer that helps both the commenter and anyone arriving later. In Groups, follow the group's published rules as well as the platform standards.

Scenario: confusion about what a free guide includes

Post: announcement for a free AI workflow guide.

Comment: "Is this just a list of tools, or does it show what to do?"

Weak draft: "Download it and find out!"

Platform-native reply: "It is built around implementation. You will choose one outcome, define the handoff between 2 tools, run a small test, and use a checklist to decide whether the workflow is worth keeping."

Why it fits Facebook: it answers the practical concern in a way that remains useful when the comment is seen outside the immediate posting window.

Check before posting: confirm that the guide delivers every item named in the reply. Do not turn a helpful answer into an unsupported sales promise.

TikTok: respond to the exact moment

TikTok comments often use shorthand that makes sense only beside the video. Reference the demonstrated moment, keep the language natural, and do not imitate slang that the brand would not use in conversation. TikTok's updated guidelines emphasize safety, authenticity, and clear community rules. Official TikTok overview

Scenario: a request for the prompt shown on screen

Video: a short demo turning messy notes into a task list.

Comment: "Drop the prompt."

Weak draft: "You already know we got you. Link in bio!"

Platform-native reply: "Use: 'Turn these notes into tasks with owner, due date, dependency, and missing information. Do not invent details.' The last sentence is what keeps the task list honest."

Why it fits TikTok: it delivers immediate value tied to the video and highlights the important instruction rather than forcing a click.

Check before posting: make sure the prompt is safe to use with the kind of notes shown. If the notes may contain personal or company data, add a privacy warning.

LinkedIn: add professional substance

LinkedIn replies work best when they advance the work-related discussion. Avoid generic praise, canned networking language, and unsupported authority claims. LinkedIn requires authentic identity, professional expression, and non-spammy participation. Official LinkedIn policies

Scenario: constructive disagreement on an operations post

Post: a recommendation to add human review before an automated email sends.

Comment: "Human approval defeats the purpose of automation."

Weak draft: "Interesting perspective. Automation is changing everything."

Platform-native reply: "For low-risk internal steps, full automation may be appropriate. We keep approval at the point where an external message, payment, or irreversible action occurs. The useful design question is where the cost of a wrong action becomes material."

Why it fits LinkedIn: it acknowledges the counterpoint, defines a decision rule, and gives other professionals something concrete to apply.

Check before posting: disclose your affiliation if you mention a product or client relationship. Remove claims about results unless they can be supported.

Reddit: contribute before you promote

Reddit is organized around individual communities with their own rules and norms. Read the subreddit rules, search for earlier answers, and disclose relevant affiliation. A useful Reddit response can be longer than a reply elsewhere, but it should solve the question without requiring a click. Reddit's platform rules and moderator expectations emphasize authentic participation and community-specific governance. Official Reddit policies

Scenario: a detailed request for workflow advice

Thread: "How do small teams review AI-written research without slowing everything down?"

Weak draft: "We made a guide about this. Check our profile."

Platform-native reply:

"Use a claim ledger between research and writing. Give each row a claim, source URL, source date, quoted evidence, and review status. The writer receives only approved rows. This is slower than accepting the first draft, but much faster than untangling unsupported claims at the end. I work on NEXAIUM, which publishes guides on this topic, but the method above does not require our tools."

Why it fits Reddit: the answer stands on its own, offers a specific method, and discloses the brand connection without making promotion the price of help.

Check before posting: confirm that brand participation is permitted in the subreddit. If the community bans promotional links, do not add one.

Step 4: generate 3 controlled options

Do not ask an AI tool for "the perfect reply." Ask for options with different conversation outcomes, then select or rewrite one.

You are drafting options for a human social media manager. Do not post, send, hide, report, or take any account action.

Use only the context packet below. Do not invent facts, customer history, policies, links, offers, or current trends. If a needed fact is missing, mark [VERIFY] instead of guessing. Never include private information.

First classify the comment as one of: answer, acknowledge, correct, support, escalate, moderate, or no reply.

Then produce:
1. Direct option: answer the point with the least possible friction.
2. Warm option: recognize the person and add one useful detail.
3. Boundary option: correct, redirect, or close without prolonging conflict.

For each option include:
- Draft reply
- Intended outcome
- Fact or link that needs verification
- Risk note

Match the named platform without stereotypes, forced slang, or unnecessary emoji. Follow the brand voice and maximum length in the context packet.

Context packet:
[PASTE COMPLETED CONTEXT PACKET]

Step 5: review with the SEND check

Use this 4-part gate before posting:

S: Source

• Is every factual claim supported by an approved source or firsthand account?
• Does a linked page prove the exact statement?
• Has a changing fact been checked recently?

E: Empathy

• Does the reply address what the person actually said?
• Is the tone calm without becoming cold or defensive?
• Would the reply still feel respectful if shown without context?

N: Necessity

• Does the reply answer, clarify, acknowledge, protect, or route?
• Is a public reply better than support, moderation, or silence?
• Can any sentence be removed without losing value?

D: Destination

• Is the next step clear and safe?
• Does a support route avoid requesting private data in public?
• Does a link help the person, or is it only promotional?

If any section fails, revise or do not post.

Escalation rules

Do not draft a normal conversational response when the comment includes:

• A credible threat, doxing, stalking, or targeted harassment.
• Self-harm language or an immediate safety concern.
• Allegations that could create legal or regulatory exposure.
• A request for medical, legal, or financial advice specific to the person.
• A security vulnerability, account compromise, or exposed credentials.
• Payment details, account identifiers, or other private customer information.
• A journalist, regulator, law-enforcement request, or crisis event.

Preserve the minimum evidence required by your policy, restrict access, and send the case to the named owner. Do not promise an outcome or deadline unless the owner has approved it.

Build a lightweight reply log

Track enough information to learn without storing unnecessary personal data.

Review 20 cases after 2 weeks. Look for repeated questions that deserve a clearer post, guide, product explanation, or support article.

Finished-state checklist

• [ ] The original post and comment are in the context packet.
• [ ] Personal information has been removed before AI processing.
• [ ] The comment has been triaged before drafting.
• [ ] The reply follows the norms of the named platform without caricature.
• [ ] Every factual claim is verified or removed.
• [ ] The reply adds value and is not generic engagement bait.
• [ ] Sensitive cases follow the escalation path.
• [ ] A human has approved the final wording and action.

Put the system to work

Start by adapting the master prompt with your own voice rules in the Prompt Library. If you want draft options from the purpose-built assistant, Open the NEXAIUM Social Media Reply Agent, then apply the SEND check before using any reply.

Verification notes

The linked platform policies were reviewed on 2026-09-04. Platform rules, enforcement, interface controls, and account capabilities can change. Recheck the official policy for each platform and follow your organization's legal, safety, privacy, and escalation requirements.