If you have been on the ManyChat community forum this week, you have seen the same kind of post over and over. Comment replies are silent. DMs are landing for some commenters but not others. The automation says it sent. The follower never got it. Welcome to April 2026.

This is not speculation. Every quote in this article is pulled directly from threads posted in the last 48 hours on ManyChat''s own community forum. We are not here to dunk on a competitor. We are here to explain what is actually breaking, because if you are paying for an automation tool and your messages are vanishing, you deserve a clear answer.

What Users Are Reporting Right Now

Here are four real complaints from real ManyChat users, all posted between April 25 and April 26, 2026:

  • Tara McKee (posted April 26): "I have been paying for ManyChat for months, while it has not been working." Her Instagram comment keyword automations stopped triggering replies entirely. Source thread
  • Rahul Aravindan (posted April 26): "It is not responding to any comment right now although it was 2 days ago. I think it is only DMing half of them as well not sure what to do." Source thread
  • olgahatters (posted April 25): "Everything was working perfectly, and then suddenly Instagram comment replies stopped working (DMs still work fine)." Source thread
  • Abirami (posted April 25): The DM is triggering perfectly when users comment, but the auto reply to comments is not showing up under the post or reel. Reply to Comment toggle is on. Keywords are right. Account is connected. The public reply just is not posting.

You are looking at four different users with four different setups, all describing variations of the same pattern. The DM half of the automation is partially landing. The public comment reply is not. Some commenters get a response. Most do not. And nobody can tell them what is going on.

The Pattern: DMs Land, Comment Replies Vanish

The thing to notice is that these are two different actions. When someone comments your trigger keyword, two separate things are supposed to happen:

  1. The public reply. Your account posts a visible reply under the comment so other viewers see engagement.
  2. The private DM. Your account sends the commenter a direct message with whatever you promised them, the link, the freebie, the next step.

Both are separate calls to Meta''s Graph API. They have different rate limits. They go through different webhook subscriptions. They can fail independently. And right now, the public reply call is the one that is breaking for most people.

Why does that matter? Because if your automation tool treats the two actions as one thing, and one of them fails, you have no idea which one. The dashboard says "sent." The follower never got the DM, or the public reply never showed up, and you cannot tell the difference. You just see angry users in your DMs asking why they did not get the link.

What Is Actually Broken

Some of what is happening this week is on Meta''s side, not ManyChat''s. Meta''s webhook subscriptions, which are the notifications Instagram sends to automation platforms whenever a comment or DM arrives, have been dropping intermittently across multiple platforms for the last two weeks. We have seen the same flakiness on our end. Anyone who tells you Meta''s API is rock solid right now is either lying or has not looked at their logs.

But here is the thing. A Meta-side problem is exactly when your automation tool''s internal recovery logic matters most. Webhook drops are normal. Rate limits are normal. Transient API errors are normal. The question is: when those things happen, does your tool quietly lose your message, or does it notice, hold onto it, and try again when conditions improve?

That is the real difference between a reliable platform and a fragile one. It is not whether failures happen. It is what happens after.

Meta''s Rate Limits, In Plain English

Before we get to recovery, let us talk about why messages get dropped in the first place. Meta''s own documentation spells out the limits:

Action Meta''s Limit
Comment-triggered DMs (private replies) 750 per hour, per Instagram account
Conversation DMs (Send API) 100 to 300 per second, throttled by engaged-user count
Public comment replies No documented hourly cap, but shared with other actions

Now picture this. You post a Reel that pops off. Two thousand people drop your trigger keyword in the comments inside an hour. The math says you cannot DM all 2,000 in an hour. Your account has a 750 per hour ceiling. Meta will start refusing your sends somewhere around the 600th attempt.

That is the moment of truth. Two thousand commenters are expecting a DM. The platform can either drop the remaining 1,400 on the floor, or it can park them and finish sending them as the rate window resets over the next couple hours.

The first option is what you see when users on the ManyChat forum post things like "only half my commenters got the DM." The second option is what we built into ChatGenius from day one.

What a Real Queue Does (Without the Engineering Jargon)

When a comment hits your post in ChatGenius, three things happen that you should care about as a user:

1. Every Comment Gets Logged Before We Touch Meta

The first thing we do, before we ask AI to write a response, before we try to send anything, is record the comment in our database. That record has a status, and the status updates as the message moves through the system: pending, processing, sent, or held.

This sounds boring, but it is the entire reason we never lose a message. If something goes wrong at any point, the record is still there. We can see it. You can see it. Nothing vanishes silently.

2. We Stay Under Meta''s Limits On Purpose

Our system tracks how many sends we have made on your behalf in the last sixty minutes. When we are getting close to Meta''s ceiling, we slow down. When we are at the ceiling, we stop sending temporarily and hold the rest of the queue.

Most importantly, we treat the public comment reply and the DM as two separate things with separate budgets. If the public reply is having a bad day on Meta''s end, the DM still goes out. If the DM hits a rate limit, the public reply still posts. One failing does not kill the other.

3. When Meta Says No, We Do Not Drop The Message

This is the piece that almost no other platform handles well. When Meta returns a rate limit error and tells us "try again in fifteen minutes," we do not throw away the message. We park it in our queue with a scheduled retry time. A separate process drains the parked queue automatically as soon as Meta says we can send again.

So if your post pops off and you blow past the 750 per hour ceiling, what happens is simple. The first 700 commenters get their DM immediately. The next 1,300 are held in the queue. Over the next few hours, as the rolling window resets, the queue drains. Every commenter eventually gets the response. Nobody is forgotten.

This is the difference between paying for an automation that works on your average day and paying for one that still works on your best day, the day your post goes viral.

What This Looks Like When Things Go Wrong

Let us run through a few real scenarios that map to what ManyChat users are reporting this week:

Scenario: 5,000 commenters in two hours

What ManyChat users are experiencing: Roughly the first 750 get the DM. The rest get nothing. The dashboard shows them as "delivered" or "pending." Nobody can tell which.

What ChatGenius does: First 700 get sent immediately. The remaining 4,300 are held with retry timestamps. Over the next several hours, the queue drains automatically. Every commenter gets their DM, just spread out across the recovery window.

Scenario: Meta has a server hiccup and rejects three messages with a 500 error

What ManyChat users are experiencing: The dashboard says sent. The user never received it. The complaint shows up in your inbox a day later.

What ChatGenius does: Short transient errors get retried automatically with backoff. The message goes through on the second or third attempt. Real rate limit signals with longer wait windows go to the parked queue. Different problem, different response.

Scenario: Your DM lands but the public comment reply never shows up

What ManyChat users are experiencing: Exactly what Olga, Abirami, and Rahul described above. Inconsistent. Confusing. No way to fix it from the user end.

What ChatGenius does: The two actions are tracked separately. If the public reply fails, it is logged with the actual error reason. If we are at the ceiling on comment replies but still under budget on DMs, we prioritize the DM, the thing your customer actually sees and acts on.

Scenario: A duplicate webhook arrives

What ManyChat users are experiencing: Some users have reported the same DM going out twice, which counts twice against their conversation limit and looks unprofessional to the recipient.

What ChatGenius does: Every record is keyed by Meta''s unique message ID. If the same webhook arrives twice, the second copy is recognized as a duplicate and ignored. Your follower gets exactly one DM.

Why We Built It This Way

We built the queue this way because Meta''s API is not stable. It has good days. It has days where Instagram private replies just stop working for thirty minutes. It has days where the rate limit window is stricter than what is documented. It has days where webhooks drop. If your automation tool was designed assuming the API always works, then on the day it does not, your customers do not get answers.

We assumed the opposite. We assumed Meta would fail at some rate. We built around it.

The cost is a more complicated platform on our side. The benefit is that during the multi-day Messenger delivery incident in February, we did not get a single client support ticket about lost messages. The queue absorbed the rough patch quietly, the way it is supposed to.

What This Is Not

This is not magic. If Meta is fully down for six hours, our queue parks your messages for six hours and then sends them. We cannot send faster than Meta allows. We are subject to the same 750 per hour ceiling as every other tool. The win is not speed. The win is that nothing is silently lost.

This also will not turn a 750 per hour ceiling into a 10,000 per hour ceiling. The platform constraints are the platform constraints. What we can do is squeeze out every legitimate send Meta allows, prioritize the messages that matter, and never throw a message away because of a transient error.

If you are running paid ads to a comment-triggered DM, the structural difference is this: a tool that drops messages on rate-limit failure has no way to recover them later, because there is no record that the failure happened. A tool with a persistent queue can replay every parked message the moment Meta opens the window again. Same Meta API, same hourly cap, very different outcome for the followers waiting on your DM.

Stop Losing DMs To Silent Failures

ChatGenius was built specifically to handle the moments when Meta gets flaky. Flat-rate pricing, AI-powered conversations, and a delivery system that does not give up on your messages.

Start Free Trial Book a Demo

What You Can Do Right Now

If you are a ManyChat user dealing with these issues this week, here are practical steps:

1. Check Your Webhook Connection

Go to your ManyChat Instagram settings and refresh your permissions. Reconnect through Meta for Business if the simple refresh does not fix it. This is the most common fix when comment automations stop firing.

2. Check Meta Side Permissions

On Instagram, go to Settings and activity, then Messages and story replies, then Message controls, then Connected tools, and confirm "Allow access to messages" is on. Meta sometimes silently revokes this.

3. Document What Is Failing

Take screenshots of your dashboard, the comments that should have triggered, and the lack of response. If you are paying during an extended outage, this documentation matters for any refund conversation.

4. Have A Backup Plan

If your business depends on Instagram DM automation, do not run on a single platform with no backup. At minimum, make sure you can export your contact data and conversation history outside your current tool.

5. Consider Alternatives

The DM automation space has more good options now than it did even a year ago. Reliability, AI-native conversations, and flat-rate pricing are no longer rare. We obviously have a bias here, ChatGenius is our product, and we built it because we ran into the same frustrations you are running into right now. But we are not the only option. The point is you have options.

The Bottom Line

Comment replies disappearing while DMs only partially land is the kind of failure that costs you trust with your audience. Someone who commented expecting a freebie and got nothing does not blame ManyChat. They blame you. They unfollow. They stop engaging with your next post. The damage compounds.

The thing that frustrates us when we read these forum threads is that this kind of failure is preventable. It is what proper queueing and retry logic is for. It is the kind of plumbing nobody pays attention to until the day it stops working, and on that day, it is the only thing that matters.

If you are dealing with comment replies not working, DMs only reaching half your commenters, or any of the other issues we covered above, and you want to talk through your options, grab a slot on our calendar. We will be honest with you about whether ChatGenius is a fit, and if it is not, we will point you somewhere that is.

Sources