The comments under an Instagram ad are full of people asking to buy: "price?", "do you ship to Ireland?", "is the medium back in stock?". Every one of those is a prospect raising their hand. Replying fast — publicly or in a DM — is how you convert them before they cool off. This guide covers how to automate those replies without losing control or breaking Meta's rules.
The two ways to reply
On an Instagram ad comment you have two response channels:
- A public reply. Visible under the comment. Good for answering a question once where the answer helps other readers too ("Yes — we ship worldwide, link in bio").
- A private reply. A one-time direct message sent in response to a public comment. This moves a buyer into a private conversation where you can take the sale further.
A private reply is powerful but governed by Meta. You may send exactly one private reply per comment, and only within a limited window after the comment is posted — Meta currently allows seven days. Miss the window and the private-reply option is gone. Automating the detection-and-reply step is what lets you stay inside that window reliably.
Detecting buyer intent
Auto-replying to everything is a bad idea — you would spam your own commenters and waste effort on praise and noise. The useful trigger is buyer intent.
ROAS Shield's classifier reads each comment and flags the ones that signal someone is close to a purchase: questions about price, availability, shipping, sizing, or how to order. These buyer-intent comments are treated as a first-class signal. They are protected from destructive moderation and routed to you (or to an automatic reply rule) to convert.
Draft-only by default — the safety rule
The single most important thing about auto-replying with AI is that ROAS Shield does not send AI replies automatically out of the box. Every AI-drafted reply defaults to draft-only: it is written, then it waits for a human to approve and send it.
Auto-send is something you opt into deliberately, and even then it only fires when several conditions are all true at once:
- the workspace and a specific rule have enabled auto-send,
- the model's confidence is at least 0.85,
- the reply is not flagged as needing a human, and
- there are no risk flags on the comment.
On top of that, regulated industries — alcohol, gambling, finance, health, CBD, pharma — are forced to draft-only regardless of settings. The point is that an automated reply should never embarrass your brand or break a platform rule because the model was unsure.
There is also a loop guard: ROAS Shield caps automatic replies to three per thread per 24 hours and never replies to your own Page's comments, so two automations cannot talk to each other in a loop.
Setting it up
In broad strokes:
- Connect your Instagram Business account (linked to a Facebook Page) to ROAS Shield through Meta's official login flow.
- Let the classifier start flagging buyer-intent comments on your ads.
- Review the drafted replies in your inbox, edit any you want, and send.
- Once you trust the drafts, optionally turn on auto-send for the rules and confidence thresholds you are comfortable with.
Everything runs over the official Graph API — no scraping or browser automation, which would put your account at risk and would never pass Meta's app review.
How this protects performance
Fast, relevant replies do two things for a paid ad: they convert the buyer who asked, and they make the comment thread look responsive to everyone else reading it. That is the same goal as hiding spam — see protecting your ROAS from ad-comment spam for the defensive half of the job, and hiding comments on Facebook ads for the manual basics.
Pricing
Drafting AI replies is available on every paid plan, starting at £19/month (Starter). Higher tiers add a knowledge base so replies are grounded in your own product facts, up to the £199/month Scale tier with 500,000 comments/month. See the pricing page and the FAQ for details.