Playbooks · 6 min read
Instagram Automation Explained, Before You Buy Anything
This is not the setup guide. If you have already decided and want the week-one sequence — connect, build one, test it on yourself, read what got blocked — that guide exists and is paced deliberately slowly.
This one is for the stage before that. You have seen a reel where someone says "comment PRICE and I'll DM you", you half understand what happened, and you are trying to work out whether any of it applies to you. Nothing below asks you to sign up for anything.
What is actually happening when someone comments PRICE
The mental model most people arrive with is that a robot is logged into the account, watching, typing. That is one of the two ways this is built, and it is the one that gets accounts restricted.
The other way — the only one worth your attention — is that the account owner authorised a tool once through Meta's own consent screen. Meta issued that tool a token. From then on, Instagram sends the tool a notification each time a new comment appears, and the tool can send a reply back through Meta's API. There is no password involved, no browser extension, no phone that has to stay switched on. It is two servers talking, with Meta's permission and Meta's logs.
That distinction is the entire safety question, and it is covered properly in its own post. For now, hold on to one sentence: an automation tool is something Meta hands a key to, not something that borrows yours.
The three surfaces, and why they behave differently
Instagram is not one channel. It is three, and they run on different clocks.
Comments on your own posts and Reels. You can reply publicly under the post, and you can send that person a direct message off the back of their comment.
Story replies and mentions. Someone answers your story, or tags you in theirs. That arrives as a message.
Direct messages. Somebody writes to you. You can write back.
There is no fourth surface. There is no "message everyone who follows me", no "message everyone who liked this", no "find people who follow my competitor". Those are not features anybody has and been quiet about. They are not permissions Meta issues.
The vocabulary
Five words carry most of the meaning. Learn these and every product page in this category becomes readable.
Trigger. The condition that starts an automation. In practice: a specific post, plus a set of keywords. Someone comments one of those words under that post, the automation wakes up. Matching normally ignores case and punctuation, so Price? and price are the same event. A negative keyword is the reverse — a word that means do not reply, and not interested is the one every account eventually needs.
Window. How long the platform lets you answer. This is the rule people find most surprising, because it is not a product limitation and no tool can widen it.
A comment can be answered for seven days. A direct message can be answered within twenty-four hours of that person's most recent message to you. Your own reply does not restart the clock; only theirs does. When the window closes, the send is refused outright, which is why a well-built tool shows you window refusals rather than pretending the message went.
Gate. The set of checks a reply has to survive before it is allowed out. Ours runs ten of them in a fixed order — kill_switch, connection, takeover, window, dedupe, cooldown, quiet_hours, rate_budget, credits, content_safety — and records the reason when one stops a send. You do not need to memorise that list. You need to know that a gate exists, that it will refuse things, and that a tool which never refuses anything is not simpler, it is less careful.
Credit. The unit of metering. In our case, a reply written from your own template is free and unlimited, because sending stored text costs nothing interesting. A credit is spent only when the AI composes something new: one credit, one reply. Which means the honest question about pricing is not "how many DMs do I send" but "how much of my inbox actually repeats".
Takeover. The automation stands down the instant a human replies to that thread by hand. This is the setting that decides whether automation feels like help or like a second person talking over you.
What none of it does
Worth saying plainly, because the category advertises loosely.
Automation replies to people. It does not create anything. No caption writing, no hashtag research, no content calendar, no scheduling, no publishing, no follower growth, no analytics dashboard. It cannot tell you which Reel will do well. And on our side specifically: no WhatsApp, no Messenger, no Telegram — one channel, Instagram.
What good looks like, so you can recognise it
You are not buying a message sender. You are buying judgement about when not to send.
Anyone can send a reply. The product is the refusal, and whether it tells you why.
Three things worth checking on any tool before the free tier runs out:
- Can you send a real reply to yourself before anyone else sees it? A test that runs the full pipeline and lands in your own inbox is the only honest preview. A rendered mockup is not.
- Can you see what it declined to send, and why? If refusals are invisible, you will not learn that your window is closing or your keyword is too broad until a customer tells you.
- Does it stop when you start typing? Takeover is a small feature that prevents the single worst outcome, which is a bot cheerfully finishing a conversation you were already having.
Where to go next
If the inbox test above came back yes, the doing is a week's work, not an afternoon's, and the week-one guide walks it with the screenshots.
If you want the argument rather than the instructions — what Meta enforces, what actually happens to accounts that push it — read the safety post first. It is the one that decides which half of this category you should even be shopping in.
And if you would rather read the concessions before the pitch, our own review lists what we are bad at first. Free tier is 100 credits, no card, which is enough to answer the only question that matters: do the replies sound like you.
