# Replying on WhatsApp is no longer free

> From October 1 Meta charges for every reply inside the 24-hour window. Seven adjustments so that it does not land on your invoice.

<!-- Fuente: https://www.shimliapp.com/en/blog/replying-on-whatsapp-is-no-longer-free · idioma: en -->

[Support](https://www.shimliapp.com/en/blog/category/atencion)

## Replying on WhatsApp is no longer free

From October 1 Meta charges for every reply inside the 24-hour window. Seven adjustments so that it does not land on your invoice.

**Luis Aguilar**
October 1, 2026
9 min read

IN SHORT

- As of October 1, 2026 Meta charges for service messages — every reply that is not a template — at the same rate as each market's utility template.
- Each phone number gets 1,000 free service messages a month, with no roll-over. For a high-volume operation that is a cushion, not a plan.
- Service does not enter the volume tiers: utility and authentication get cheaper with scale, service stays at list rate.
- An unadjusted flow bills six messages where an adjusted one bills two: same case, same answer, a third of the cost.

CONTENTS

- What is charged now
- The four rules that set your invoice
- What a reply costs
- Seven adjustments so you do not overpay
- The same case, two flows
- The metric that just changed price
- The underlying read

For two years a WhatsApp conversation had a flat price. You paid to open it and
whatever came next was free: three messages or forty.

As of today you pay per message. That turns the length of your conversations
into a budget line, and most of what can be done about it is configuration, not
redesign.

### What is charged now

| Criterion | Through September 30 | From October 1 |
| --- | --- | --- |
| An agent's reply in the 24 h window | Free | Charged at the market's utility rate |
| Utility template in reply to the customer | Free | Charged |
| Free allowance of service messages | Did not exist: all were free | 1,000 a month per number, no roll-over |
| Volume discount on service | Did not apply | Still does not: service is charged at list rate |

The last row is the one most often missed: service is the only category that does not get cheaper with volume, and it is precisely the one that grows with the length of the conversation.

The mechanics of when you can send did not move. The customer service window
still lasts 24 hours and still resets every time the customer writes. Reactions
— the emoji on a message — are still free and do not even draw from the
allowance. Governments and non-profits Meta recognises as eligible keep service
free through December 31, 2027.

There is a second window worth not confusing with the service one, and this
change left it untouched. When a customer arrives through a Click-to-WhatsApp ad
or a Facebook Page call-to-action button and the business replies within the
first 24 hours, that reply is not charged and opens a **72-hour free entry point window** : while it runs, no message to that customer is billed, including
templates that normally are. It decides what is charged, not what can be sent: a
service message still needs the 24-hour window open.

### The four rules that set your invoice

Every lever below comes out of these four.

**You are charged per delivered message, not per conversation.** The unit of cost
stopped being the case and became the bubble.

**Service has no volume discount.** Meta’s tiers lower the price of utility and
authentication as monthly sending grows. Service is not in that table: the one
category that grows with the length of your conversations is the one you cannot
negotiate down by scale.

**The 1,000 free are per number, not per account,** and they are a single
allowance shared between 1:1 and group sends. For an operation handling thousands
of cases a month they are gone in a day: a cushion for someone starting out, not
a lever for an institution.

**With no payment method, delivery stops.** Meta delivers what fits inside the
allowance and stops delivering the rest. No invoice arrives later: support goes
down at message 1,001.

The other side of the conversation. Every reply leaving a floor like this one, inside the 24-hour window, stopped being free on October 1.

### What a reply costs

| Colombia | $0.0008 |
| --- | --- |
| United States and Canada | $0.0034 |
| Brazil | $0.0068 |
| Mexico | $0.0085 |
| Central America, Ecuador and Dominican Rep. “Rest of Latin America” on the rate card | $0.0113 |
| Chile | $0.0200 |
| Argentina | $0.0260 |
| Peru went up 50% on this same October 1 | $0.0300 |

Central America does not exist as its own market on the rate card: Honduras, Guatemala, Nicaragua, El Salvador, Costa Rica and Panama fall under “Rest of Latin America”, at $0.0113. That is fourteen times Colombia's rate and higher than Mexico's or Brazil's. Only two rates went up in the region on October 1: utility and authentication in Peru, from $0.0200 to $0.0300, and marketing in Mexico, from $0.0305 to $0.0397.
Meta's official rate card, effective October 1, 2026

The range runs 1 to 37 between Colombia and Peru. A regional operation treating
its WhatsApp cost as a single number is averaging markets that differ by an
order of magnitude.

It is worth looking at the other column too. In Mexico a marketing message costs
$0.0397 and a service message $0.0085: **4.7 times more** . The category each
message goes out under weighs more than any volume negotiation.

### Seven adjustments so you do not overpay

None of them requires changing provider or rebuilding the operation. They run
from most to least impact, and the first four are configured in the agent.

#### In the agent

**1. One reply, one message.** Bubbling is the most profitable adjustment and the
easiest: a greeting, a clarification and the answer in three bubbles are three
billed messages where one would do. If your agent is set to “reply naturally” by
splitting text, that now triples the cost of every turn.

**2. Ask for all the data at once.** An agent that asks for the ID, then the
account number, then the date spends three messages and three waits. A single
message asking for all three resolves the same and costs a third. With buttons or
a list where possible: every misunderstanding is two more turns.

**3. Cut the messages that do nothing.** “One moment please”, “thanks for
writing”, “anything else I can help with?”. They do not resolve, do not inform
and are now billed. In an agent this is an explicit rule in its policy, not a
style preference.

**4. Acknowledge with a reaction.** It is the only service message type still
free, and it does not draw from the allowance either. Where a “✔ received” goes
out as a message today, a reaction does the same job for nothing.

#### In the templates

**5. Review every template’s category.** Meta states in writing that the assigned
category is the business’s responsibility, and that using a template means
accepting the charges of whatever category it carries at that moment. A template
that should be utility but is classified as marketing costs almost five times
more in Mexico. Worth one pass over the full list.

**6. Do not push to a template what fits in the window.** With the 24-hour window
open, replying with a service message costs the utility rate. Replying with a
marketing template costs the marketing rate. Same content, same conversation,
four or five times the difference.

#### In the operation

**7. Shift spend to Click-to-WhatsApp and load the payment method.**
Conversations that start from an ad, or from a Facebook Page call-to-action
button, open a 72-hour free entry point window in which nothing is billed. It is
the only large surface left unbilled. And the payment method is not
administrative: it is the one item on this list that can leave an operation
unable to reply.

### The same case, two flows

A balance enquiry handled by an agent, message by message

#### Unadjusted flow

- Customer: Hi, I need my account balance
- Business: Hi there! Thanks for writing to us 👋 Charged · message 1
- Business: Could you confirm your ID number? Charged · message 2
- Customer: 0801-1990-04567
- Business: Thank you. And the account number? Charged · message 3
- Customer: 3300-4521-8899
- Business: One moment please Charged · message 4
- Business: Your available balance is L 12,480.00 Charged · message 5
- Business: Anything else I can help with? Charged · message 6

**6**
messages billed

#### Adjusted flow

- Customer: Hi, I need my account balance
- Business: Happy to help. To look it up I need your ID number and your account number. Charged · message 1
- Customer: 0801-1990-04567 / 3300-4521-8899
- Business: Your available balance is L 12,480.00, as of 14:20 today. If you need the transaction detail, tell me the date range. Charged · message 2
- Reaction: 👍 on the customer's message Not charged

**2**
messages billed

Same case, same answer. What changes is how many times something is sent. In Central America, at $0.0113 per message, that difference is four and a half cents per enquiry: across 20,000 enquiries a month, about $900 a month.
That is 4 fewer messages per case, with the same answer.

That last figure is arithmetic, not a promise: four fewer messages per case,
multiplied by your market’s rate and by your monthly cases. The
[WhatsApp calculator](https://www.shimliapp.com/en/calculadora-whatsapp) runs it with your own numbers,
with the free allowance and the ads window already deducted.

### The metric that just changed price

While the window was free, the question for an operation was **how many conversations can I open** . From today it is **how many messages does it take me to close one** .

Almost nobody has that second one on a dashboard, because until yesterday it cost
nothing. It is the first thing worth instrumenting: without it, the seven
adjustments above are applied blind and there is no way to tell which one worked.
The [walkthrough of a support operation](https://www.shimliapp.com/en/casos/atencion) shows where it gets
measured, from the enquiry coming in to the case closed.

Four questions before your next committee

- How many messages does it take you to close a case today, and how many of those resolve nothing? It is the figure that moves the channel's cost the most and the one almost nobody measures.
- Does your agent split replies across several bubbles? If it does, every reply costs two or three times as much as of today, and it is fixed in the configuration.
- When did someone last review your templates' categories? One classified as marketing costs almost five times one classified as utility.
- Does your cost dashboard read the webhook's billable field? That field changed value today, so as of today it is under-counting.

[Work out what you will pay at your volume →](https://www.shimliapp.com/en/calculadora-whatsapp)

### The underlying read

Meta spent two years giving away the expensive part of the conversation so that
businesses would move to the channel. The move already happened, and today the
subsidy ends.

For an institution running volume that is neither bad news nor good: it is a
change of incentive. Until yesterday a long conversation and a short one cost the
same, which is why nobody separated them. From today the short conversation — the
one that resolves in two messages instead of six — went from being a courtesy to
the customer to being a line in the budget.

### Questions on this topic

What counts as a service message on WhatsApp? Any message your company sends that is not a template: an agent's reply, an AI agent's reply, an image, a document. They can only be sent inside the 24-hour customer service window, which opens and resets each time the customer writes. Through September 30, 2026 they were not charged; from October 1 they are.

What can I adjust in my AI agent to pay less? Four things, in order of impact: have every reply go out as a single message rather than split into bubbles; have it ask for all the data it needs at once rather than one field at a time; stop it sending filler such as “one moment please” or “anything else?”; and acknowledge receipt with a reaction, which is not charged and does not draw from the free allowance.

Are the 1,000 free messages per account or per phone number? Per business phone number, and it is a single allowance shared between 1:1 and group sends. A 1:1 delivery uses one unit; a group send uses one per delivered recipient. It does not roll over from one month to the next: every month it resets to 1,000.

Do service messages get cheaper with volume? No. Meta's volume tiers apply to utility and authentication templates, not to service. An institution sending millions of templates a month earns a tier rate on utility, but every service message beyond the first 1,000 is charged at list rate no matter how many it sends.

What happens if the account has no payment method on file? Meta delivers the service messages that fit inside the free tier and stops delivering the rest. It is not an invoice that arrives later: support goes down at message 1,001. The payment method is set in the account's Billing Hub.

Useful to someone on your team?
Share it with whoever decides on this process.

[WhatsApp](https://wa.me/?text=Replying%20on%20WhatsApp%20is%20no%20longer%20free%20https%3A%2F%2Fwww.shimliapp.com%2Fen%2Fblog%2Freplying-on-whatsapp-is-no-longer-free) [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.shimliapp.com%2Fen%2Fblog%2Freplying-on-whatsapp-is-no-longer-free) [Email](mailto:?subject=Replying%20on%20WhatsApp%20is%20no%20longer%20free&body=Replying%20on%20WhatsApp%20is%20no%20longer%20free%20https%3A%2F%2Fwww.shimliapp.com%2Fen%2Fblog%2Freplying-on-whatsapp-is-no-longer-free)

Copy link

### Keep reading

[See all →](https://www.shimliapp.com/en/blog)

- Support #### Your next banking channel is already installed: WhatsApp AI agents are doing to apps what apps did to websites. What that means for an institution running real volume, and what separates a chatbot from an agent. September 25, 2026
- Support #### 80% of your queries do not need a person The problem in a support operation is not volume: it is that the 80% shares a queue with the 20% that does need judgement, and delays exactly what matters. September 15, 2026
- Platform #### FINNOSUMMIT 2026: one in three sessions was about AI Instead of summarising impressions, I counted the sessions: 22 of the 67 had AI in the title. That says a lot about where banking in the region stands. September 25, 2026

### Put an agent to work on this process.

Create the account and start today. Or we take 30 minutes and define which metric you'll move.

[Create account](https://shimli.app/user/auth/create-account)
Book a demo

[See the full support case →](https://www.shimliapp.com/en/casos/atencion)
