KewbotIn short
From October 1, 2026, Meta charges again for WhatsApp service messages: the free-form replies you send a customer inside the 24-hour window. Today they're free (they have been since November 2024). This is not a new tax or a trick: it's a return to a charge that already existed before. The 24-hour window does not change. The only thing that changes is that those replies now get billed per message, at the same rate Meta already charges for utility and authentication templates in your country. With a well-designed conversation, the real impact is small and controllable.
What does NOT change (so you can breathe)
- The 24-hour window stays the same: it opens and resets every time the customer messages you.
- Your number, your flows, and your templates keep working exactly as they do now.
- Utility and authentication templates are billed just as they were: no surprises there.
- You don't have to migrate anything or rebuild your operation. It's a cost adjustment, not a technical change.
In mid-2026 Meta announced a set of price changes on the WhatsApp Business platform. The one that made the most noise is this: service replies, which cost nothing today, get a price again. The concern is understandable, especially in support-heavy, high-volume operations. But it's worth separating the noise from the facts: the change is specific, it's documented, and you can plan for it. Let's go through it piece by piece.
The free-form reply inside the 24h window
When a customer messages you, a 24-hour service window opens. During that window, everything you reply in free-form text (a person on your team, or a bot that isn't Meta's own native agent) is a service message. That's what's free today and becomes chargeable in October. Don't confuse it with templates, which are the messages you start yourself outside the window and which were already billed.
If the customer messages you and you reply within the day, that reply used to be free. From October 1, that reply carries a per-message cost. What you start yourself with a template (reminders, notices, campaigns) was already billed and stays the same.
At your country's utility rate, per message
A service message is billed at the same per-message rate Meta already applies to utility and authentication templates in the customer's country. That number varies by market. Meta committed to publishing the final per-country rate card before September 1, 2026, so you'll be able to confirm the exact figure ahead of time.
Unlike utility and authentication templates, service messages have no volume discount: the rate is flat whether you send few or many in a month. That's why the real cost lever isn't negotiating the unit price, it's sending fewer unnecessary messages. We cover that below.
Fewer scattered messages, more resolving conversations
Here's the actionable part, and the one that separates a tidy operation from one that bleeds money. If your team (or your bot) takes 40 messages to resolve something that resolves in 6, that inefficiency stops costing only time and starts costing money. The way to contain the spend isn't to reply less: it's to reply better.
Consolidate replies: instead of five little messages in a row, one clear and complete one. Ask for several details at once instead of one at a time. Classify and route quickly so conversations don't drag on forever. And automate the repetitive with precise logic, so every exchange actually moves forward instead of going in circles. A well-designed conversation sends fewer billable messages without the customer feeling worse served. If anything, they're served faster.
Meta's native AI agent is billed separately
In parallel, from August 1, 2026, Meta started charging for its own AI agent (the Meta Business Agent), billed per token consumed rather than per message. It's a separate charge from service messages. Worth keeping on the radar so you don't mix up concepts when you read your invoice.
A service message sent by a person or a third-party tool is billed as a service message (flat per-message rate). A message generated by Meta's native agent is billed per token. Those are two different lines in your cost. Knowing which is which saves you from surprises when you interpret the numbers.
⚠️ How to read this without over-worrying
Yes, a cost that used to be zero stops being zero, and in high-volume operations you'll notice it. But it's a predictable, plannable change, with a rate card published ahead of time and clear levers to contain it. Operations that already run on tidy conversations and precise automation will barely feel it. The ones that send messages haphazardly today are the ones that will notice most, and those have plenty of room to improve. In other words: it's a good moment to review how you're conversing, not to panic.
Frequently asked questions
When does the charge start and how much is it?
From October 1, 2026. The per-message price is the same one Meta already charges for utility and authentication templates in the customer's country. The final per-country rate card is published before September 1, 2026, so you'll be able to confirm the exact figure before it starts.
Is it a new tax or something just made up now?
Neither. It's a return to a charge that already existed: Meta billed for these replies until November 2024, then made them free, and now brings the charge back. The 24-hour window and the rest of how things work don't change.
Do I have to change my number or migrate anything?
No. It's a cost adjustment, not a technical change. Your number, your flows, and your templates stay exactly the same. The only thing worth reviewing is how many service messages your operation sends and whether conversations can be made more resolving.
How do I lower the impact on my invoice?
By sending fewer unnecessary messages: consolidated replies instead of many scattered ones, asking for several details at once, classifying and routing quickly, and automating the repetitive with precise logic. It's not about serving less, it's about making every exchange move forward. A well-designed conversation costs less and also serves faster.
Should I just leave the WhatsApp API then?
For operations with real volume, hardly. WhatsApp is still where your customers are, and the API's capabilities (automation, templates, campaigns, multiple agents) are still worth far more than this cost adds when it's well managed. The key is reaching October with a tuned operation, and there a good partner is an investment that pays for itself.
Want to reach October with a tuned operation?
We help you tidy up your conversations so every message counts, with precise automation and content always under your control. Get in touch and we'll build a plan sized to your volume.
Go to Kewbot →