WhatsApp API Pricing: Your UK Guide to 2026 Changes & New Service Message Charges

If you're still calculating your cost per conversation on the official WhatsApp API, you're using a metric Meta retired in 2025. And from 1st October 2026, the new metric gets more expensive: replies your team sends within the 24-hour window will no longer be free. The deadline is fast approaching. This post explains what's changing, how much each message category will cost, and where to cut expenses without compromising customer service.

TL;DR

• Since 1st July 2025, Meta charges per message delivered, no longer per 24-hour window. The cost per conversation has ceased to exist as a billing unit.
• From 1st October 2026, service messages (free replies within the customer service window) and utility templates sent within the window will be charged.
• Anyone without a valid payment method registered by 30th September 2026 will stop delivering service messages.
• For example, in Brazil, marketing messages cost around R$0.31 per message; utility and authentication messages around R$0.035. Service messages will now cost the same as utility messages.
• Since July 2026, new Brazilian accounts are invoiced in Brazilian Reals, with invoices issued by a local entity. Dollar-denominated accounts will stop delivering messages in Brazil from 1st July 2027.

I've worked with WhatsApp API integrations since the platform opened to official providers, and I track every Meta pricing revision because they change the product's design, not just the spreadsheet. What follows is what I'd tell a client with their September invoice in hand.

Why Cost Per Conversation Died in 2025 and the Billing Model Changed

Until June 2025, Meta charged per conversation: a 24-hour window opened with the first message charged, and within it, you could send as many messages as you wished for the same price. The cost per conversation was predictable — simply multiply the number of customer interactions by the category value to get your monthly bill.

On 1st July 2025, this model ended. Meta's official pricing documentation began describing charges per message delivered, with the tariff defined by two variables: the recipient's country and the template category (marketing, utility, or authentication).

The practical difference is significant. Under the old model, three utility templates for the same client on the same day cost one conversation. Under the new model, they cost three messages. Workflows that stacked templates — confirmation, reminder, payment link — tripled in price without anyone touching a line of code. This is why so many companies looked at their 2026 invoice and didn't recognise the figure: the volume didn't increase, the billing unit simply shrank.

Why Cost Per Conversation Still Appears in Proposals

The term has survived in providers' commercial vocabulary because it helps compare plans, and cost per conversation remains the metric managers understand. For internal budgeting, however, change the question: instead of "how much does a conversation cost?", ask how many charged messages each customer journey consumes. This is the count Meta bills for.

The Five Dates in 2026 That Will Impact Your Cost Per Conversation

2026 has been the year of the most significant pricing adjustments since the Cloud API opened. It's worth having these dates pinned to your monitor:

Date	What Changes	Invoice Impact
1st June 2026	Meta releases WhatsApp Business account currency migration APIs	Neutral — enables switch to BRL (Brazilian Reals)
1st July 2026	New Brazilian accounts are created in Brazilian Reals, invoiced by local entity	End of exchange rate risk; local BRL invoicing
1st August 2026	Meta Business Agents begin charging per token (US$2 per 1 million)	New cost for those using Meta's native AI
1st September 2026	Meta publishes official service message tariffs by country	Allows budgeting for October with actual figures
1st October 2026	Service messages and utility templates sent within the window become chargeable	Direct increase for inbound operations

The last date is the one that hurts, as it redefines the cost per conversation for those operating inbound services. Until 30th September 2026, replying to a customer within the window is free; the very next day, it's not. For a support operation with 40,000 replies per month, this represents a new line on the spreadsheet that previously showed zero.

There's also an operational detail that catches out the unwary: according to providers who have already communicated the change, anyone without a valid payment method registered by 30th September 2026 will stop delivering service messages. It's not a fine or a warning — it's a delivery block. Check today if billing is active on your account.

How Much Each Message Category Costs in Brazil Today (Example)

To arrive at the actual cost per conversation for your operation, start with the tariff for each category. Brazilian tariffs are among the cheapest in the world for utility and authentication, and mid-range for marketing. The values below use an exchange rate reference of R$5.00 per US dollar and reflect the tables in circulation in September 2026:

Category	Initiator	Order of Magnitude (BR)	When Charged
Marketing	Business	~ R$0.31 per message	Always. No volume discount
Utility	Business	~ R$0.035 per message	Outside the window today; inside it too from 01/10/2026
Authentication	Business	~ R$0.035 per message	OTP and two-factor verification
Service	Customer	R$0.00 until 30/09/2026	Thereafter: same as utility tariff

Treat these figures as an order of magnitude and confirm with Meta's official rate card: the table changes by country, by revision round, and, in the Brazilian case, also by exchange rate.

Note the asymmetry: marketing costs almost ten times more than utility. This is the most valuable insight in this entire post. Most of the waste I find during audits doesn't come from volume — it comes from templates classified in the wrong category.

The Category Trap

Meta classifies the template upon approval, looking at the content. "Your order 1234 is out for delivery" is utility. The same text with "and enjoy 10% off your next purchase" becomes marketing — the promotional phrase added by a well-intentioned growth team multiplies the cost of that send by nine.

Before approving a new template, ask yourself: would this message exist if the customer hadn't bought anything? If the answer is no, it's a utility message and needs to be written as such — no offers, no commercial call to action, no promotional emojis.

The 24-Hour Window: What Ends on 1st October 2026

The customer service window opens when a user sends a message to your number and lasts 24 hours. Within this window, today, you can reply freely, without a template and without cost. This is what sustains the low cost per conversation for any WhatsApp support operation.

On 1st October 2026, this free period ends at two points:

1. Service messages — free replies sent by a human agent, chatbot, or third-party AI within the window opened by the customer.
2. Utility templates sent within the window — which were free in this context since November 2024 and will now be charged again.

Both will now cost the same as the utility and authentication tariff for the recipient's country. In Brazil, for example, this means around R$0.035 per message delivered — cheap per unit, but expensive at scale. The cost per conversation for an operation that exchanges 12 messages per interaction and handles 5,000 monthly interactions will go from R$0 to approximately R$2,100 per month.

What Remains Free

Not everything has become a cost. Two exemptions remain in place and are invaluable:

• 72-hour Free Entry Point — when the customer initiates the conversation by clicking a Click to WhatsApp ad or a Facebook page CTA button, the entire exchange remains free for 72 hours. This is the only path that has become relatively cheaper.
• Monthly service message allowance — providers like 360dialog and YCloud have communicated an allowance of 1,000 free service messages per phone number, per month, with no rollover to the next month. Confirm this with your provider, as Meta has not highlighted this allowance with the same fanfare in public announcements.

This allowance changes the calculation for small businesses: those sending fewer than a thousand replies per number per month will, in practice, continue to pay nothing.

Service Messages: What Will Be Charged and What Remains Free

It's worth detailing what Meta considers a service message, because the definition is broader than it seems. It is any free-form message, outside of a template, that the business sends within a window opened by the customer. It doesn't matter who typed it, and this has a direct effect on your operation's cost per conversation:

• Human agent replying in the inbox — charged
• Flow-based chatbot replying automatically — charged
• Third-party AI agent (Claude, GPT, Gemini) replying — charged
• Automatic 'we're out of hours' message — charged
• Read receipts, typing indicators, reactions — not messages, not counted

The size of the message doesn't change the price. A three-word reply costs the same as an entire paragraph. The design consequence is immediate: fragmented messages have become a financial error. That friendly chatbot that sends 'Hi!', then 'I'm Ana, your virtual assistant', then 'How can I help?' has just tripled the cost of that interaction.

In September 2026, the Brazilian specialised press reported the change prominently — the Mobile Time announced the confirmation of reply charges and their equivalence with the utility tariff. This isn't a vendor rumour; it's published policy.

Meta Business Agents: The Native AI That Charges Per Token

In parallel, Meta launched Meta Business Agents, its native AI agent infrastructure across WhatsApp, Messenger, and Instagram. The free testing period ran from 1st to 31st July 2026. Since 1st August 2026, charges are per token: US$2 per 1 million tokens, with message delivery already embedded in this price.

In practice, a typical reply consumes between 20,000 and 25,000 tokens, which works out to between 4 and 5 US cents per message — approximately R$0.20 to R$0.25. Compare this with the R$0.035 for a regular service message: the cost per conversation with a native agent is five to seven times higher than handling it with your own agent.

When the Native Agent Pays Off

This isn't a condemnation. Meta Business Agents make sense when you don't have a technical team, want something live in days, and have low volume. There's a relevant exemption: replies from agents created on the Meta platform are exempt from service message charges, as they are already paid for by the token.

When it doesn't pay off: volume above a few thousand conversations per month, the need to control the model, or any scenario where you already have your own API key. In such cases, connecting your own model to a platform running on the official API is much cheaper — and you maintain control of the prompt, history, and data.

Invoicing in Brazilian Reals: The Migration Almost No One Has Made

This is the silent item of 2026 and will catch many off guard. Since 1st July 2026, every new WhatsApp Business account with a billing address in Brazil is invoiced in Brazilian Reals, with an invoice issued by Meta's local entity in the country. From 16th July 2026, all eligible providers can create accounts in BRL.

What this solves: an end to month-to-month exchange rate exposure and payment via local methods, instead of international dollar invoices. What this requires: those with older dollar-denominated accounts need to migrate. Meta released currency migration APIs on 1st June 2026, reducing the process to two calls — and stipulated that, from 1st July 2027, foreign currency accounts will stop delivering messages to users in Brazil. If your account predates July 2026, put this migration on your roadmap now: it's not urgent today, but it will be next June along with everyone else.

Seven Leaks That Inflate Your Bill and How to Plug Them

When auditing real operations, the same patterns emerge. The seven leaks below account for most of the avoidable cost per conversation I encounter, and none of them require switching platforms to resolve:

1. Utility template with a promotional phrase. This gets classified as marketing upon approval and costs almost nine times more. Rewrite and resubmit in the correct category.
2. Chatbot that fragments replies. Each bubble is a charged message from October onwards. Combine everything into a single turn, using line breaks.
3. Mass sending without database hygiene. Messages delivered to inactive numbers are still charged. Cleanse contacts that haven't interacted in over 180 days.
4. Ignoring the Free Entry Point. Click to WhatsApp campaigns open up 72 free hours. Sending traffic directly to a WhatsApp link loses the entire exemption.
5. Collecting data via ping-pong. Five questions become five charged messages. A WhatsApp Flow collects everything in one form, in a single message.
6. Re-engaging via marketing when utility would suffice. An overdue bill reminder is a utility message. Only use marketing when the message is genuinely promotional.
7. Single number for everything. The service message allowance is per number. Separating support from sales into distinct numbers multiplies your monthly allowance — but requires a plan per number, so do the maths first.

If you want to understand why the official API remains the right path even with these tariffs, it's worth reading our comparison between the WhatsApp Business App and the official API and the complete guide to the WhatsApp Cloud API.

How Agathas Web Addresses This

The part of the cost per conversation you control is the volume of charged messages. The part you don't control is Meta's tariff — unless someone is adding a margin on top of it. And this is more common than it should be: a good portion of WhatsApp platforms on the market resell Meta's messages with embedded markup and without showing the breakdown. We've already written about this hidden cost of message markup.

Our Voyia customer service platform, built on the official API, was founded on the opposite premise: full pass-through of Meta's tariffs, without a penny of margin. You pay the monthly plan fee for the software and pay Meta for what Meta charges. What's included:

• Integrated official Meta API, with a verified number and templates submitted by our team
• Multi-agent inbox with queues, agent transfers, and quick replies — reducing unnecessary messages per interaction
• AI integration in a BYO (Bring Your Own key) model: connect Claude, GPT, Gemini, or DeepSeek with your own API key. You pay the model provider directly, with no intermediary, avoiding the per-token cost of Meta's native agent.
• No-code chatbot to automate what currently consumes agent messages
• Mass messaging with approved templates and tag-based segmentation
• Delivery and read reports to see exactly where charged volume is concentrated
• iOS and Android App, operation compliant with UK GDPR and Data Protection Act 2018, and support in Portuguese

In practice, implementation begins with an audit: we examine your workflows, count how many charged messages each journey consumes, and rewrite what's necessary before activating your number. We've seen operations cut a third of their monthly sends just by fixing template categories and unifying fragmented replies.

Those who prefer to evaluate before committing can view plans, scope, and Meta's message pass-through table directly on the Voyia page — a table we update as the official rate card changes, including to reflect service message charges from October. If your case is more complex than a standard implementation — ERP to integrate, multiple units, regulated workflow — our consulting team will design the solution before you commit.

Conclusion: Cost Per Conversation is Now an Architectural Decision

In 2024, your WhatsApp bill was a procurement item: you negotiated the plan and went about your business. In 2026, it's an engineering item. Every extra message in a flow, every misclassified template, and every chatty chatbot will appear on your bill at the end of the month.

The deadline is short. By 30th September 2026, confirm your payment method on your account to avoid stopping service message delivery. By 1st October, review template categories and unify fragmented replies. By mid-2027, migrate your account to Brazilian Reals. Once this is done, your cost per conversation stops being a matter of luck and becomes a project.

If you want to conduct this review with a partner who passes through Meta's tariffs without margin and shows you the breakdown line by line, discover Voyia and speak to our team. The first conversation is a diagnostic of your current workflow — and often pays for itself in the first invoice after adjustment.