Instagram DM automations

Say "comment LINK" in your caption. Someone comments LINK. Zippy DMs them your link, optionally replies under their comment, and tags them.

That part is table stakes — ManyChat has done it for years. Here's the part that isn't: the link in the DM is always a Zippy link, so the tap springs the real native app instead of Instagram's in-app browser, and the app-open lands in your stats. A comment-to-DM tool's report ends at the tap. This one carries on past it — and no other DM tool can act on that outcome per person, sending the follower who got stranded last time different words and a different link.

DM automations are a Legend feature.

How do I auto-DM people who comment on my Instagram post?

Once you're connected, in Instagram → New automation:

1. Pick the trigger. "Someone comments on my post" is the 90% case.
2. Set the keywords. LINK, PRICES, SEND IT — up to 10, matched case-insensitively. Choose contains (the comment includes one of your words) or is exactly (the whole comment is one of them). Leave the keywords empty and it fires on every comment, which is a real setting and not a mistake.
3. Optionally pin it to one post so a rule only listens to a single reel.
4. Write the DM. Just the words — see below for why you don't type the link.
5. Attach the link. Pick one of your Zippy links from the dropdown.
6. Optionally add a public reply ("check your DMs 💌") and tags for everyone the rule replies to.

The builder shows you a live preview of the exact message your follower receives and a one-line summary of when it fires, both rendered by the same code the server uses — so what you see cannot drift from what gets sent.

Up to 25 rules per connected account.

What are the four triggers?

All four use the same keyword matching and the same DM composer. The difference is which Instagram surface starts them, and — importantly — which reply window Instagram gives you.

Why can't I type a link into the DM text?

Because a raw URL in a DM is exactly the leak this feature exists to plug.

Tap a bare https://… inside Instagram and it opens in Instagram's in-app browser: a webview where your follower is logged into nothing, Apple Pay doesn't exist, and cookies are walled off from the pixel you're paying for. The DM landed. The checkout didn't. And the tap reports nothing back to you.

So the link is attached, not typed. You pick a Zippy link; Zippy appends its short URL to your message at send time. If you paste a raw URL into the DM text anyway, saving the rule fails with a message explaining why, and even if a rule somehow got saved that way the send itself refuses. That's deliberate: we'd rather skip one DM than ship one that can't open the app and can't be measured.

The Deep Link Debugger shows you the difference on your own destination, side by side.

When will Instagram not let my automation send?

Instagram restricts unsolicited messaging hard, and an account it restricts is not something anyone can hand back to you. So Zippy refuses to send whenever the rules are unclear, and tells you rather than failing quietly.

Every skip is a row in the automation's history with that explanation in plain English — never a raw Meta error code, never silence.

One structural guarantee worth stating: every DM Zippy sends is a reply to a specific person who just interacted with you. There is no way to point this at a list. There is no audience field to fill in. That isn't a policy we promise to follow, it's the shape of the thing.

How do I know if my DM automation actually worked?

Two numbers, and the second one is the whole point.

Sends — how many DMs went out, how many were skipped, how many failed, and how many distinct people you reached. Every comment-to-DM tool has this.

App opens — of the people who tapped the link in that DM, how many landed in the real native app versus the in-app browser versus nowhere useful. Zippy's redirect engine records what happened after the tap, so a rule's report reads:

• opened — they reached the real app. Logged in, cards saved, pixel alive.
• browser — the webview won.
• broken — the hand-off went nowhere useful.

That's the answer to "my DM automation gets clicks but no sales", and it's the number a click counter structurally cannot produce.

App-open rates are a sampled estimate and are labelled as such, and they are iOS-only — an Android tap is recorded as unmeasured and counted in neither the top nor the bottom of the rate. A rule whose link has no measured taps yet reports no rate at all rather than a misleading 0%.

Can I see the whole funnel, from the comment to the opened app?

Yes — four steps, and only the first two exist in any other comment-to-DM tool:

Every rival's funnel stops at step 2, because the link they send is a plain URL that dies in Instagram's in-app browser and reports nothing back. Steps 3 and 4 exist here only because the DM carries an engine-backed Zippy link.

The gap between step 3 and step 4 is the number to stare at. Those are people who wanted what you offered enough to tap, and then got stranded in a logged-out webview with no saved cards and no pixel. Zippy prints that count in plain words underneath the funnel, because it is the audience you already paid to reach and then lost at the door.

What the funnel will not do

• It will not pad itself. A step Zippy cannot measure reads unavailable, never 0. Three things cause that, and each says which: the rule sends no link, analytics isn't connected, or the rule sends personal links whose taps we can't total yet.
• It will not invent a drop-off across a step it couldn't measure. If step 3 is unavailable, there is no "lost between 3 and 4" figure.
• A brand-new rule gets a sentence, not a funnel of zeros.
• Steps 1–2 and steps 3–4 are not the same kind of number and are never presented as one ledger. The first two are exact database counts; the last two are sampled and extrapolated, marked sampled on the step itself, and always shown as whole numbers — a decimal there would be false precision.

Can I share it?

There's a one-tap share card: a single image drawing all four steps, with the count of people who opened the real app as the headline. Nobody else can post that picture.

The card is all-or-nothing on purpose. It renders only when all four steps are real numbers and somebody actually opened the app; otherwise you get a message explaining which step is missing. An image travels without the page that framed it, so a funnel drawn with a hole in it would read to strangers as a complete one. The sampling note is printed on the card itself for the same reason.

Can I do more with the people who replied?

Yes — that's what the tags are for. A rule stamps its tags on everyone it replies to, so "everyone who commented DROP on the launch reel" becomes a group you can act on rather than a number you watched go by.

From there:

• Audience segments slice the resulting taps by country, device, and outcome — a segment filtered to opened is your true-fan list.
• Email capture turns those taps into contacts you own, off Instagram entirely. Instagram can restrict your reach tomorrow; an email list can't be restricted by anyone.
• Webhooks push app.opened / app.browser / app.broken to your own stack the moment a DM tap resolves — no polling.

How do I say something different to people who didn't open the app?

Add a condition. This is the one thing on this page that no other DM tool can do.

Every Zippy link reports whether the tap sprang the real native app or died in Instagram's in-app browser. Because we already know that, per person, a rule can act on it:

If they tapped last time but got stuck in the in-app browser → send them the store link instead, with different words.

ManyChat can branch on a tag. It can't branch on whether the app actually launched, because the raw URL it sends reports nothing back. We're not the only tool that measures app opens — URL Genius reports an app-arrival percentage too — we're the one that lets a DM rule act on it per person.

In the builder, under Say it differently to some people, add up to 3 lines. Each line is one condition, one message, and optionally a different link. They're read top to bottom, first match wins, and everyone the lines miss gets your main message.

What can I branch on?

The app-open conditions only apply to links we sent that person directly. Public taps by anyone else stay anonymous — see the identity join.

Which condition worked?

Open Results on the rule. Under the headline numbers there's a row per condition — how many people took it, how many people that was, and that branch's own app-open rate.

That last column is the point. If your "stuck in the browser" line sends a different link, you can see whether it actually lands those people in the app more often than the one it replaced. Everyone the conditions missed shows up as Everyone else, so the rows always add up to the rule's total.

Rates are whole numbers from sampled click data over 90 days, same as everywhere else in Zippy. A branch that's never fired shows as empty, not as 0%.

Do not act on a few points of difference between two branches. The underlying signal is a page-visibility inference with known false readings — a phone lock mid-launch reads as opened, a cold start slower than nine seconds reads as browser — the counts are sampled, and every rate here is an iOS rate, because Android taps are unmeasured and sit in neither the top nor the bottom of it. Compare branches with a clear gap, never a decimal. How the number is made and where it lies.

Buttons

Your link can ride as words — Instagram turns it into a tap — or it can be a button. A button is one tap instead of two, and it's the only way to find out which offer people took.

Under 5. Add buttons, each button is three fields: what it does, what it says, and (for a link button) which of your links it opens. The builder lets you add three, which is Instagram's cap on the shape a link button has to use — see the two shapes for when the real ceiling is higher.

Labels stop at 20 characters, because that's where Instagram cuts them off.

Why do my buttons look different sometimes?

Because Instagram has two different controls, and a message can only ever use one of them. Zippy picks for you and never mixes the two, so what you build is what goes out.

Both numbers are Instagram's, not ours. Three is all a button template holds; thirteen is all one message can carry in quick replies. Zippy stores up to 13 buttons on a rule and applies whichever ceiling the shape it picked actually has.

Two consequences worth knowing before you build a rule:

• A long message drops your link buttons. The template's text tops out at 640 characters, but a DM can be 1,000. Write past 640 and Zippy sends your full words rather than truncating them — which means the quick-reply shape, which holds no link buttons. Your link is still in the message, because it always is.
• The email ask and a link button never appear together. A button template can't carry an email ask at all, so if the message went out as a template, the ask is dropped. Put the email ask on a rule with no link button and it ships as a quick reply.

What happens when someone taps a quick reply?

• A tag button stamps that person with the button's own words, lowercased — "Send me the guide" becomes the tag send me the guide. Instagram hands us back a payload we wrote ourselves, so a chip from somebody else's tool can't tag your people by accident. Your next automation can then answer them with the A tag they already have condition.
• The email ask hands over the address they picked, and it lands in Audience. Instagram writes this button's label itself and forbids us from setting one, so the label you type on it is only there to help you tell your buttons apart — your follower never sees it. That's also why the consent wording sits in the message body rather than on the button: 20 characters is nowhere near enough room for words someone can meaningfully agree to, and Instagram allows no label here anyway.

Either way the tap arrives as an ordinary message in the thread, so it also opens Instagram's 24-hour reply window — the same window every other rule of yours works inside.

Two places buttons don't appear

Both are Instagram's limits, not ours, and neither one can leave someone stuck:

• On desktop. Instagram doesn't draw buttons in its web DMs at all.
• On the DM that answers a comment. Instagram only documents plain text there, so we send plain text. Buttons work on every other trigger — keyword, story reply, story mention — because those all open the standard 24-hour reply window.

In both cases your link is still in the words. If your rule has its own link, that's the one. If the only link you attached is on a button, we put that one in the message too. So the worst case is that someone reads a message and taps a link, which is exactly what happened before buttons existed.

Which button opened the app?

Open Results on the rule. Under the per-condition table there's a row per button — its label, what kind it is, and its own app-open rate.

This is the column a tap counter can't give you. Two buttons can get the same number of taps and land wildly different numbers of people inside the real app, because they point at different destinations. A button that carries no link (a tag button, or the email ask) shows as empty rather than 0% — it has nothing to measure, and pretending otherwise would be a lie.

Same honesty rules as everywhere: whole numbers, sampled clicks, 90 days, iOS only (Android taps are unmeasured and score in nothing), and a heuristic underneath. Compare buttons with a clear gap, never a decimal.

Notes

• Legend plan. Downgrade and your rules go read-only, never deleted. Links you already DM'd keep redirecting forever.
• Nothing is sent to your followers while our Meta app is unapproved — see setup.
• Same rules, both surfaces. Web and mobile read the same trigger copy, the same preview, and the same skip explanations from one source, so they can't tell you two different things.
• Your token is never readable back, not even once. See how Zippy stores your Instagram access.

Related

• Instagram setup — the prerequisites, the permissions, and the honest approval status.
• ManyChat — the version that works today, on any plan, with zero approval.
• Deep Link Debugger — watch a link escape the in-app browser before you attach it to a rule.
• Webhooks · n8n — push the outcome into your own automation.
• Audience segments · Email capture — what to do with the audience a rule builds.
• UTM builder · A/B split testing — attribute and race the links your DMs carry.
• Blog: ManyChat DMs a link that dies in the in-app browser
• Blog: Branch on whether the app actually opened
• Blog: Every button is a measured spring
• Blog: The funnel that ends at the opened app
• Changelog: Instagram DM automations and keyword, story reply & story mention triggers and conditions that branch on app-open and buttons with per-button app-open rates and the comment-to-app-open funnel