Charged per conversation or per message? The fine print that changes your bill

The short answer

It depends which bill you are looking at, because there are two. If you are asking whether WhatsApp is charged per conversation or per message, Meta's answer today is per message. Your platform's answer can be anything else, and that is where two quotes that look alike are not.

Meta has charged per delivered message since 1 July 2025. The conversation model, the one built on 24-hour windows, is deprecated.

The two bills

Nobody sells you "WhatsApp". They sell you two things that arrive together.

	Who charges	What it counts	When it shows up
Meta layer	Meta, against your account	Delivered messages	Monthly, based on what you sent
Platform layer	Your provider	Depends on the provider	Flat subscription, or usage

The first is the same for everybody: nobody negotiates Meta's rates. The second is where each provider picks its unit, and where you have to read.

What Meta actually counts

It counts messages you deliver, and only when they are templates.

Until 1 July 2025 the model was different: a 24-hour conversation opened and that whole window was billed. Meta's documentation now marks that model as deprecated. If a quote mentions "1,000 free conversations a month", it was copied from an old page.

Today the customer writes to you, a 24-hour window opens, and replying inside that window costs you nothing.

What changes on 1 October 2026

That last part stops being true. From that date Meta begins charging for service messages and utility templates sent inside the 24-hour window.

It is still per message. Five messages are five charges.

What your provider counts

There is no standard here, which is why the question is worth asking bluntly: what unit are you counting?

The three you will run into:

• Per conversation. Every customer who writes consumes one unit, whether you reply once or twenty times. Rewards short conversations.
• Per message. Every message counts. A long conversation gets expensive even when it is a single customer.
• Per AI reply. It only counts when the AI answers. What a person on your team types costs nothing.

At Alumbra, conversations have no cap. What gets measured are AI replies, and there is one counter per organization.

One detail worth knowing before the invoice rather than after: that counter is shared. If you also use the assistant inside the panel, its turns come out of the same pool as the replies to your customers.

Why the unit changes your bill

Back to a real case. A business we serve handles about 312 conversations and 4,137 messages a month: 1,247 inbound and 2,890 outbound.

That same operation produces very different numbers under the three models:

• Per conversation: 312 units.
• Per message: 2,890 outbound units.
• Per AI reply: roughly half the conversations resolve without a person stepping in, so consumption looks closer to the first number than the second.

Same operation. The gap between 312 and 2,890 is the unit, not the usage.

What to ask before you sign

1. What unit are you counting, and how many are included?
2. Do my team's own messages consume, or only the AI's?
3. What happens when I go over? Cut off, charged, or warned?
4. Is Meta's cost separate or included?

The third gets asked least and hurts most. At Alumbra there are warnings at 75% and 100%, and the cutoff only lands at 110%. When it cuts, the AI pauses and conversations stay open so your team can keep replying by hand.

Key takeaways

• Meta has charged per delivered message since July 2025. The per-conversation model is deprecated.
• Any table promising "1,000 free conversations a month" comes from the old model.
• On 1 October 2026, service messages and utility templates inside the 24-hour window start being charged.
• The platform layer counts whatever each provider decides: conversations, messages or AI replies.
• The same operation can read as 312 or 2,890 units depending on the unit. Ask which one before comparing prices.